Art, community and AI: images for an affective memory

This research examines the potential of artificial intelligence (AI) image generation technologies for humanistic applications, specifically focusing on the creation, reconstruction, and reinvention of nonexistent visual archives through which communities can restore affective and identity-based memories. The theoretical framework draws from multiple perspectives: affect theory of postmemory, techno-aesthetic and affective apparatus theory, the concept of insubordination of signs, distribution of the sensible, and theories of affective memory and visual archives. Through case study analysis, this investigation examines the creative, ethical, and theoretical-methodological strategies employed in four artistic and experimental projects, each driven by distinct aesthetic, emotional, political, activist, and vindicatory objectives. The research identifies three distinct levels of intervention in fractured memory through AI: the socio-political (evidenced in projects addressing historical invisibility and state violence), the community-cultural (manifested in projects recovering cultural traditions), and the therapeutic-personal (focused on individual memory restoration). The findings demonstrate the significant potential for expanding generative AI applications toward emotional repair, memory reinvention, and transformation of established sentiment structures, introducing the original concepts of “algorithmic postmemory” as a framework for understanding AI’s active role in memory construction and “affective symbolic documentalism” for comprehending the testimonial value of AI-generated imagery. These applications challenge dominant discourses that render specific social collectives invisible within society while offering new methodological approaches for the restoration of fractured memories and historical healing.

1 Introduction

In recent years, the rapid advancement of artificial intelligence (AI) technologies has opened new horizons across various fields, including arts and humanities. This research explores the intersection between AI, art, and affective memory, examining how generative AI models can be leveraged to create, reconstruct, and reinvent visual archives that serve humanistic purposes. Our study focuses on the potential of AI-generated images to restore and reimagine affective and identity-based memories of diverse communities, particularly those whose histories have been marginalized or erased.

The intersection between artificial intelligence and affects is an emerging field. Some applications of AI in the field of affects or emotions include promoting emotional states, for example, affective generative music created for entertainment or health purposes, with the clear aim of influencing the listener’s emotional state (Dash and Agres, 2024) or chatbots called emotional support systems that interact with humans using affective technologies to detect emotions, decide and simulate affective responses (Devillers, 2021). Additionally, there are communicative and artistic practices by users with different types of generative artificial intelligence applications. This is the case of image creation related to affective memory.

It is crucial to recognize visual production created with AI. Image creators through generative AI software have developed various artistic styles. Furthermore, prompt design and prompt templates have also become a creative practice. “Glitches” have served as inspiration for the development of new styles (Chang et al., 2023). However, representation biases that exist when there is an uncritical use of generative AI cannot be ignored. Fernández de Caleya Vázquez and Garrido-Merchán (2024) emphasize that if the image data used in the models is biased towards variables such as gender or skin color, the generated content can perpetuate harmful stereotypes for vulnerable, historically discriminated or invisible groups, such as those related to physical features, sexual preferences or disability.

Artistic production has the potential to function as a critical-artistic artifact. The uses proposed by ordinary people, artists, social groups, or activists demonstrate the ways in which society receives technologies and what it is interested in expressing with and about them. This research is situated at the intersection of three aspects: artificial intelligence, visual communication, and affects. This research examines the creative, ethical, and theoretical-methodological strategies employed in four Latin American visual projects based on generative AI: Un archivo queer inexistente, 2022 (Rivas San Martín, 2022a); Exhumar la Memor.IA (Séptimo, 2023); IAbuelas, 2023 (Barros, 2023); Synthetic Memories, 2023 (Domestic Data Streamers, 2023). Each has been driven by distinct aesthetic, emotional, political, activist, and vindicatory objectives. It highlights the humanistic potential of AI in reconstructing lost memory and affective relationships through images that recover both a lost “this has been” and a generative “this is” (Barthes, 1984), deploying artificial symbolic production in service of inclusive and restorative affective memory.

This research examines how generative AI technologies can be leveraged to create, reconstruct, and reinvent visual archives that restore affective and identity-based memories for communities whose histories have been marginalized or erased. Specifically, we investigate: (1) How can AI models function as techno-aesthetic apparatuses that enable the restoration of fractured memories and affective relationships? and (2) What creative, ethical, and theoretical strategies prove most effective when using AI for memory restoration across different contexts (socio-political, community-cultural, and therapeutic-personal)?

We are interested in entering into the study of selected cases and guided by these questions, making a series of inquiries about how this philosophical horizon of reflection on art and visuality becomes specific categories of analysis about image and artificial intelligence. We are particularly interested in transferring these theoretical tools from the critical tradition of art to this new object of study and reflection to create a bridge between the new images produced by generative intelligences and the general discussion about affectivity in contemporary art. From this connection will emerge the specific questions that are formulated at the time of the methodological proposal, based on the cases and made with a hermeneutical spirit, to then review the coincidences and divergences between these examples.

The theoretical framework of this research draws from a multidisciplinary approach, combining perspectives from postmemory (Hirsch, 1997, 2008, 2012), digital mediation of memory (Van Dijck, 2007), techno-aesthetic apparatus theory (Simondon, 2007; Déotte, 2012), affective artifacts (Belk, 2013; Piredda, 2020), the concept of insubordination of signs (Richard, 1994, 2011), the distribution of the sensible (Rancière, 2009), and the theory of affective memory and visual archive (Didi-Huberman, 2012; Barthes, 1984). This interdisciplinary approach allows us to critically examine the creative, ethical, and theoretical-methodological strategies employed in four pioneering artistic and experimental projects: Rivas San Martín (2022a), Séptimo (2023), IAbuelas (2023), and Domestic Data Streamers (2023).

The research aims to construct a theoretical-methodological-analytical framework that allows understanding the various visual expressions of AI from a humanistic perspective, with special emphasis on affective memory and emotional support. Additionally, this study aims to thoroughly analyze the creative processes of the selected cases, all produced in the Ibero-American context, which adds a specific cultural dimension to our research.

A fundamental aspect of our approach is conceptualizing generative image models as techno-aesthetic apparatuses (Simondon, 2007; Déotte, 2012). This perspective allows us to explore how these technological systems can construct a space of symbiosis, a sensible world, and a narrative hybridization between humans and machines, transcending mere technical functionality to enter the terrain of aesthetic and affective experience.

In parallel, our research delves into the meaning of visual signs as tools of humanistic “insubordination” (Richard, 1994). This approach is particularly relevant when examining how vulnerable communities appropriate these tools to counter the dominant and, at times, exclusionary use of image and AI in the globalized socioeconomic landscape. The political subversion of symbolic production of generative AI images by marginalized communities not only opens an unprecedented horizon of humanistic possibilities but also fosters a new distribution of the visual sensible (Rancière, 2009), reconfiguring power relations in the field of representation.

Our research addresses the crucial task of reflecting on the construction of fractured affective memory (Didi-Huberman, 2012; Barthes, 1984). In this context, we analyze the importance of visual symbolic construction processes in the formation and recovery of vulnerable or violated identities. This aspect becomes especially relevant when considering communities affected by forced disappearance, stigmatized by their gender condition, or requiring an affective restoration of memory to recognize and reaffirm themselves. AI, in this sense, presents itself as a potential tool for the reconstruction and healing of these collective and individual memories.

Through the analysis of these case studies through the prism of our theoretical framework, we seek to unravel the humanistic potential of AI in expanding emotional repair, memory reinvention, and transformation of established sentiment structures. This research contributes to the growing body of work exploring the intersection between technology and human experience, offering new perspectives on how AI can be used as a tool for social inclusion, historical vindication, and emotional healing.

To analyze the humanistic possibilities of AI in restoring affective memory, we build a theoretical framework that integrates postmemory theory and digital mediation, the conceptualization of AI models as techno-aesthetic apparatuses and affective artifacts, the principle of sign insubordination in relation to active memory and cultural resistance, and the analysis of how AI fosters a new distribution of the visual sensible. From this framework, we will develop a methodology with specific categories of analysis to examine four pioneering projects in depth, exploring their objectives, creative strategies, processes, community involvement, ethical considerations, and contributions to affective memory.

2 Postmemory, mediation, and digital restoration of fractured memory

The concept of postmemory enables an understanding of the intergenerational transmission of historical trauma. Hirsch (1997) defines postmemory as the relationship that a subsequent generation develops with the personal, collective, and cultural trauma of those who preceded them. The connection to the past is not established through direct memory, but through imaginative investment and creation mediated through stories, behaviors, and images transmitted within the family and social environment. In later works, Hirsch (2008) argues that these experiences can be transmitted so deeply and affectively that they seem to constitute memories of their own.

Postmemory can be constructed through two processes that Assmann (2009) distinguishes: communicative memory and cultural memory. On one hand, communicative memory operates through direct family communication, oral tradition, and face-to-face narratives. On the other hand, cultural memory is realized through cultural artifacts, such as monuments and institutional representations. Traumatic experiences are transmitted through these two spheres, the private and the public. Both contribute to the construction of shared historical narratives and the formation of collective identities.

In At Memory’s Edge, Young (2000) complements these perspectives by analyzing the material and artistic manifestations of postmemory. Cultural representations, monuments, contemporary artworks, and memorials serve as links for the intergenerational transmission of trauma. Young’s analysis of the concept of “mediated memory” helps understand how cultural and artistic manifestations contribute to the preservation and construction of collective memory, especially in relation to the traumatic events of the Holocaust.

Moreover, in the Latin American context, Sarlo (2005) formulates a constructive perspective on the concept of postmemory, without setting aside the critical vision. In Past Time: Memory Culture and Subjective Turn, the author examines the application of the concept in the context of Latin American dictatorships, recognizing its usefulness for analyzing the processes of trauma transmission in post-dictatorial societies. Sarlo emphasizes the need to consider local particularities and the different ways in which societies process and transmit their traumatic experiences.

Critical perspective of Derrida’s (1997) has been fundamental in understanding how postmemory transforms the very nature of the archive. Derrida conceives it as a dynamic space in constant reconfiguration, where the unconscious intentions and desires of those who create and manipulate the archive intertwine with physical documents. Foucault (1983) complements this vision by analyzing how postmemory challenges traditional documentation systems and the power structures that sustain them. His analysis of “disciplinary power” reveals how error and distortion can function as critical tools that allow the emergence of silenced voices.

With the advent of the digital age, Van Dijck (2007) introduces the concept of “mediated memories” to analyze how new technologies are reconfiguring the intergenerational transmission of trauma. Their work examines how digitization not only modifies the storage of memories but fundamentally transforms the processes of cultural memory transmission.

Drozdzewski and Birdsall (2019) expand this line of research by exploring the digital, spatial, and material “turns” in memory studies. Work of Garde-Hansen’s (2011) delves into the social dimension of this digital transformation, analyzing how social networks and digital platforms have become new spaces for constructing and preserving collective memory. Their research reveals how these new forms of archive are revolutionizing our relationship with the past. Erll et al.'s (2009) contribution on “remediation” is fundamental to understanding how digital media not only preserves but actively transforms existing memories. This process creates new layers of meaning and forms of interaction with the past, redefining our understanding of collective memory.

The emergence of generative artificial intelligence marks a new chapter in this evolution. Shobeiri and Westgeest (2024) concept of “hypothetical memories” provides a framework for understanding these new forms of synthetic memory. AI models not only archive but can recreate and reinterpret memorial narratives, raising fundamental questions about authenticity and algorithmic mediation of memory.

From an approach integrating psychoanalytic theory (Freud, 1980, 1992) and archival studies (Sekula, 1986; Buchloh, 1999; Guasch, 2011), Martín Prada examines how contemporary artists such as Rivas San Martín, Mavropoulou, Séptimo, and Colectivo 8,552 employ generative AI to develop a new “archive aesthetics.” This approach challenges hegemonic visual narratives and the relationship between image, memory, and temporality through both the creation of fictional photographic archives that reveal historical absences and the critical revision of photoconceptual practices (Martín Prada, 2025).

From the Barthesian perspective, the photographic “this has been” becomes the generative “this is.” The symbolic and historical value of the image is constructed through its emotional and affective referents. The photographic image acts as a trigger for affective memory, as a “synthetic punctum,” connecting the viewer with their past and feelings, beyond the referent. The category of “punctum,” articulated by Barthes (1984), emerges here as a key element for understanding the emotional and affective production detonated symbolically, through the image as a vehicle of affective memory.

Perspective of Didi-Huberman’s (2012) on the image as vestige acquires new relevance in this digital context. His theorization about how images can help revive past experiences, even those artificially generated, suggests new pathways for understanding and restoring fractured memory in the digital age. The “synthetic vestiges” generated by AI, although not direct historical evidence, provide symbolic and affective value that opens new possibilities for exploring and understanding the past.

The critical-artistic use of AI for the symbolic production of memory brings us closer to what Didi-Huberman (2012) formulated about the importance of image in fractured memory, where each image is a vestige that can help revive past experiences, despite the destruction of historical archives and censorship. In the case of AI, we are dealing with “synthetic vestiges,” which are not evidence, but do provide symbolic marking with affective and historical value that goes beyond the indexical. For Didi-Huberman (2012), the archive is “almost always grayish” due to the losses and destructions that surround it, but by discovering the “memory of fire” in the images that have survived, a connection with the past can be restored. Images are not, in this case, simple cuts of reality, but are traces that synthesize diverse and complex times, which have the power to make visible experiences that inscribe memory in history, making them fundamental tools for understanding and restoring fractured memory.

A fundamental concept that runs through these discussions is that of “archontic violence,” developed by Derrida (1997), which refers to the violence exercised by the guardians of the archive (the Archons) over collective memory. This violence manifests in institutional control over how memory is stored, organized, and accessed, and in the imposition of “official” interpretations that silence alternative voices. Postmemory, in this context, functions as a form of resistance that questions these power structures, allowing the emergence of marginalized narratives and subaltern experiences that have been systematically excluded from official archives.

3 Generative AI image models as techno-aesthetic apparatuses and affective artifacts

The case studies in this research have been created with generative image models using AI. The systems used in each case have allowed authors to reconstruct images and memories that never existed, created through information synthesis processes.

In the contemporary scenario, we are witnessing the emergence of large generative AI models for producing image, sound, or text. Image generative models like Stable Diffusion, DALL-E 3 (OpenAI), or Midjourney (among many others), or text “large language models” like ChatGPT4 or Claude. These AI models are large artificial intelligence systems designed to process and generate information, also called transformers, diffusion models, Large Language Models (LLM), Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs). These systems are all examples of deep learning architectures.

In this research, we conceptualize these generative AI models as techno-aesthetic apparatuses from Simondon’s (2007) and Déotte (2012) theoretical perspective. These “apparatuses” are evolving technical entities that transcend their purely instrumental function. They are models that can be conceived as “technical individuals” that evolve through processes of concretization, mediating between human beings and nature, adapting to their informational environment. Furthermore, they have a transindividual potential and can generate new forms of collective connection, expression, and communication, reconfiguring relationships between individuals and groups, from the philosophical perspective of Simondon’s (2007).

From the perspective of Déotte (2012) and Flusser (2006, 2011), these models have the potential to configure and reconfigure the sensitivity and perception of a reality, world, or epoch by acting as techno-aesthetic apparatuses. They are symbolic production apparatuses that foster forms of inscription and materialization of ideas and concepts, altering our relationship with cultural production. These techno-aesthetic apparatuses not only generate content but also transform relationships between objects, signs, and human beings.

The techno-aesthetic apparatuses redefine our experience of time in relation to artistic creation and knowledge production, establishing new paradigms of perception and expression. They mediate our experience of the world, are not neutral, and configure sensitivity and perception in specific ways. They perform a production of the sensible, not only reproducing reality but having the potential to produce new forms of sensitivity, aesthetic experience, and historical perception of an epoch (Déotte, 2012).

The relationship between affective artifacts, techno-aesthetic apparatuses, and generative artificial intelligence opens new horizons for understanding how digital objects modify and configure our affective experience. Belk (2013) and Piredda (2020) note that social media profiles can function as affective artifacts in the digital context; new generative AI technologies introduce an additional dimension by allowing the creation and manipulation of affective memories that never really existed.

The emergence of generative AI, as a digital and cultural artifact, once again highlights the possibility of generating archives that can symbolically produce emotional bonds (Pataranutaporn et al., 2023). The ability to generate texts, sound, still or moving images, remediating photographic or cinematographic discourses of moments that never occurred, or the voice of deceased loved ones, significantly expands the concept of “affective artifact” studied by Piredda (2020), beyond the preservation of existing memories towards the active creation of new affective experiences.

The technological capability to generate synthetic affective artifacts reconfigures the sociocultural and historical relationship with memory, identity, and the “extended self” described by Belk (2013), raising a new horizon about the role of postmemory in the emotional and political resignification of the historical archive and the role of technological mediation in the construction of collective and individual identities.

4 The insubordination of signs, art, active memory, and resistance

Through critical use, techno-aesthetic apparatuses are capable of performing a symbolic production of “insubordinate” signs through art. According to Richard (1994), fractured, forgotten, or invisible identities can be recovered through a critical use of art and signs, with the aim of healing historical traumas derived from human rights violations. The key concept that Richard proposes in response to the phenomena of violence that occurred during the Chilean dictatorship (1973–1990), applicable to other similar processes, is that of the “insubordination of signs” as a form of resistance. This notion encompasses processes that, through artistic expressions, help understand and catalyze historical healing. The insubordination of signs invites questioning and challenging meaning structures in culture, promoting a critical and dynamic vision of social and artistic reality.

Insubordinate signs, through art, foster a reinvention of memory that facilitates the elaboration of unfinished mourning and generate an active memory through non-linear narratives. These narratives evidence the crises of meaning and vindicate individual and collective identities that have been wounded or forgotten. Likewise, these signs constitute a form of cultural resistance to hegemony, making singularities visible in the face of the hypervisibility of what is socially accepted.

This approach develops an aesthetic of rupture and a politics of difference, enabling symbolic and narrative processes of destruction, reconstruction, and deconstruction of dominant discourses. Ultimately, the insubordination of signs contributes to a critical reinterpretation of the past and the construction of new historical narratives that challenge hegemonic versions of individual or collective memory (Richard, 1994).

From this perspective, critical artistic manifestations contribute to the reinvention of individual or collective memory in traumatic or post-traumatic contexts. The concept of active memory that Richard (1994) formulates allows us to think about a dynamic process in the construction and reconstruction of collective memory in a participatory way. This concept establishes possible connections between past and present, functioning as a tool for understanding and critiquing current reality. Equally, it acts as a form of resistance against hegemonic official narratives and challenges simplified versions of history, giving voice to marginalized experiences.

Active memory entails a continuous reinterpretation of history, resignifying events and experiences to achieve a more complex understanding of identity and culture. Richard emphasizes the fundamental role of cultural and artistic practices in this process, as they allow exploring the tensions between forgetting and remembering, thus maintaining a living and relevant memory in the contemporary context.

Art understood from this perspective becomes a powerful tool of resistance against oppression, officialized rhetoric, invisibilization, or systematic forgetting. Critical artistic practices can transcend the aesthetic, providing a political and social commitment that addresses memory, identity, and trauma. Art understood in this way allows questioning dominant narratives, making injustices visible, and reconfiguring oppressed identities, especially in contexts of repression. Critical art’s role is fundamental in preserving historical memory and creating “scenarios of resistance” that promote dialogue and the collective construction of forms of struggle (Richard, 1994, 2013). Art thus becomes an ideal medium for articulating critical reflection against processes of oppression, forgetting, and hegemonic discourses.

5 AI and an emerging distribution of the visual sensible

The symbolic visual production generated through AI is subject to the historical uses of a techno-aesthetic apparatus that has the capacity to reconfigure relationships between subjects, objects, and machines, and therefore, to reconfigure the sensory perception of the world and what is conceived as reality. The distribution of the sensible is intrinsically related to the emergence of techno-aesthetic devices and their critical or uncritical use.

In this sense, symbolic production machines are pouring out rhetoricized or poetic representations that affect what Rancière (2009) called the distribution of the sensible. This “distribution of the sensible” defines how experiences, perceptions, and visibilities are distributed within a community, implying that there is a specific organization of what is visible (and audible), as well as forms of participation and exclusion in the political and aesthetic realm.

For Rancière (2009), the distribution of the sensible not only determines who can see or be seen or who can speak or be heard but also conditions how identities are formed and how power relationships are configured. Thus, the distribution of the sensible is a political matter related to the possibility of being recognized, made visible, or having a voice in public space.

Aesthetics in this context, the critical-artistic use of images for example, has a fundamental role in this political-sensible distribution, as artistic practices can challenge and reconfigure established norms about what is sensible and what is not, thus opening new possibilities for experience, visibility, the construction of identities, the reconfiguration of historical memory, and the participation of oppressed, marginalized, or forgotten communities in public spaces (Rancière, 2009).

6 Method

This research adopts a mixed methodological approach, combining case study selection with qualitative analysis, close reading, and in-depth examination. This methodological design allows us to comprehensively examine the humanistic applications of generative AI models in the context of affective postmemory, artistic production, and digital algorithmic mediation.

The sample of case studies has been selected following a systematic qualitative methodology that allowed identifying them within the larger scenario of the evolution of humanistic and non-humanistic manifestations of AI (Bañuelos, 2022). The selection process was conducted using the following specific criteria: (1) Projects using generative AI specifically for affective memory restoration, with documented methodologies and accessible outcomes; (2) Projects representing diverse memory contexts (socio-political trauma, community cultural memory, and personal memory loss); (3) Projects belonging to the Ibero-American sphere, which allows us to examine the cultural and social particularities of this context; and (4) Projects created between 2022 and 2023, coinciding with the widespread availability of advanced generative AI tools like DALL-E, Midjourney, and Stable Diffusion. This temporal and geographical delimitation allows us to focus on projects that are not only relevant to our theoretical framework but also represent different contemporary approaches to using AI for memory restoration.

To analyze the selected cases, we have developed an analytical methodology incorporating the key concepts of our theoretical framework. This methodology is structured around seven main elements of analysis: objective, strategy, and creative process, use of AI, community involvement, ethical construction, humanistic values, and affective memory.

In the objective element, we examine each project’s declared and underlying purposes, evaluating how they align with the restoration of affective memory and humanistic uses of AI. The categories of strategy and creative process, and use of AI, allow us to analyze the specific AI techniques employed, the creative processes involved, and how these relate to the concepts of techno-aesthetic apparatuses (Simondon, 2007; Déotte, 2012). When evaluating community involvement, we consider how each project integrates and engages affected communities, examining participation, co-creation, and technological appropriation aspects.

Ethical construction and humanistic values are analyzed considering the concept of insubordination of signs (Richard, 1994) and the distribution of the sensible (Rancière, 2009), examining the ethical considerations and values that guide each project. Finally, in the affective memory category, we analyze how each project contributes to the construction, reconstruction, or strengthening of affective memory, using the theoretical perspectives of Didi-Huberman (2012) and Barthes (1984).

Data collection is carried out through a combination of methods, including documentary analysis of materials produced by each project, observation, and analysis of the artistic and visual works generated. Data analysis follows a systematic, heuristic, and categorization process, using the previously defined categories of analysis as an initial guide while allowing the emergence of new themes.

Data collection was carried out through a comprehensive documentary analysis of materials produced by each project, including artistic manifestos, published interviews with creators, exhibition documentation, and the generated images. For each case, 15–20 representative images and their accompanying texts were identified and analyzed. The interpretative procedure followed a hermeneutic analysis in three phases: (1) detailed description of the formal and technical characteristics of the generated images; (2) contextualization of these characteristics within the established theoretical framework; and (3) identification of patterns and emerging themes through comparative analysis between the different cases.

To systematize the analysis of selected cases, a methodological protocol was developed that integrates documentary analysis with the study of visual productions generated through AI. This protocol is structured from an analytical matrix that articulates central theoretical categories with specific elements of analysis in each case, allowing a systematic approach to the object of study.

The analytical matrix is organized in three fundamental dimensions. The first, techno-aesthetic and artifactual, examines image generation’s technical and procedural aspects, including the specific characteristics of AI models used, prompt engineering strategies employed, and the relationship established with pre-existing photographic archives. The second dimension, memorial and affective, analyzes memory activation mechanisms and emotional connection strategies, considering both individual and collective processes, as well as implemented community validation methods. The third dimension, ethical-political, evaluates community participation processes, consent mechanisms, and specific ethical considerations of each project.

The analytical matrix (Table 1) systematizes the comparative analysis of the four cases studied through seven key conceptual categories: Actor, Cause, Artistic-Technical-Symbolic-Affective Process, Affective Memory and AI, Critical and Ethical Use of Art, Insubordination of Signs, and Distribution of the Sensible. This tool allows visualizing both the particularities of each project and the common patterns in the humanistic use of AI for the restoration of affective memory (see Table 1).

Table 1

Table 1. Analytical matrix of cases generating affective memories through AI, 2024.

This study is limited to cases from the Ibero-American context during 2022–2023, which restricts generalizability to other cultural contexts. While the selected timeframe captures the emergence of sophisticated generative AI applications, it necessarily excludes ongoing developments. Additionally, our analysis focuses on the creative and theoretical dimensions rather than quantitative measures of community impact or therapeutic efficacy, which would require longer-term studies. The study intentionally delimits its scope to artistic and experimental projects rather than clinical applications, focusing on how these technologies are being appropriated for memory restoration rather than their technical implementation details.

7 Analysis of cases

7.1 Un archivo queer inexistente (2022)

The project Un archivo queer inexistente (2022)¹ by Felipe Rivas San Martín represents a significant intervention at the intersection between generative AI, collective memory, and historical vindication of marginalized identities. This project uses generative AI as a tool to create a fictional photographic archive documenting the Latin American queer experience of the early 20th century, directly addressing the historical absence of visual records of these communities.

The project’s objective emerges from what Richard (1994) calls the “insubordination of signs,” by creating a queer counter-archive that challenges the archontic violence described by Derrida (1997). As the artist himself states: “one of the consequences of heteronormative culture is that the queer experience of the past could not leave any record or archive” (Rivas San Martín, 2022a). This denial of the archive is particularly dramatic for people from the Global South and working classes, configuring what Sarlo (2005) identifies as a double marginalization: by sexual identity and socioeconomic condition.

The project emerges from the artist’s previous work in the University Collective of Sexual Dissidence (CUDS) and his research on the intersection between queer activism, archive politics, technology, and decoloniality. As documented in the Barcelona City Council’s announcement about the exhibitions (Ajuntament de Barcelona, 2024), the project seeks to reimagine “portraits of homosexual couples, lesbian and queer or non-binary people from the early 20th century in Latin America.” The photographs were exhibited at Zona Maco and later at the Centro Cultural Clavijero, where the artist explained that “the images are plausible at first glance, but when you start looking at the details you notice these imperfections, which reveal that something does not quite fit” (Góchez, 2023). This tension between plausibility and artificiality is fundamental to the project, as Rivas San Martín (2022b) points out in his text Technopoetics of Activism and Dissidence, his work emerges from the intersection between queer critique, archive, and technology, seeking “an algorithmic critique of the queer” that questions dominant narratives about the past.

The project’s creative strategy and technical process are grounded in the use of generative AI as a techno-aesthetic apparatus, in the sense proposed by Déotte (2012). The artist uses generative AI systems to create photographs that simulate the style and aesthetics of the early 20th century, but showing homosexual couples, lesbians, and queer or non-binary people in situations of intimacy and affection. As Rivas San Martín notes, this is “a retrofuturistic exercise that uses computational algorithms to reimagine our local queer past and reclaim an archive that could not even exist” (Rivas San Martín, 2022a).

The process involves what the artist calls a “minority prompt,” a strategy of specifically designed instructions to counteract the inherent biases in the databases used to train AI models. This approach represents what Rancière (2009) calls a redistribution of the sensible, by creating new forms of visibility and representation for historically excluded identities. Rivas San Martín used Stable Diffusion 1.4, deliberately incorporating its technical limitations as aesthetic elements that reveal the artificiality of the images, while leveraging its ability to emulate Latin American historical photographic styles.

Community involvement manifests in the political and affective dimension of the project. The generated images function as what Piredda (2020) calls “affective artifacts,” allowing the Latin American LGBTQ+ community to visualize and connect with a history that was denied to them. The project activates what Van Dijck (2007) calls “mediated memories” through AI, creating new spaces for the construction and preservation of queer collective memory.

Regarding ethical construction and humanistic values, the project directly addresses what Didi-Huberman (2012) describes as the reconstruction of fractured memory. The technical imperfections in the AI-generated images, far from being a defect, function as ethical markers that evidence their constructed nature, what the artist describes as “profoundly queer imperfections of a technology still in development” (Rivas San Martín, 2022a).

The affective memory dimension of the project is articulated through what Hirsch (1997) calls postmemory, creating an imaginative connection with a past that could not be documented. The AI-generated photographs act as what Barthes (1984) would call “punctum,” in this case a “synthetic punctum” that provokes an emotional response connecting the viewer with a silenced collective history. As Young (2000) suggests, these representations serve as links for the intergenerational transmission of marginalized experiences.

The project also exemplifies what Shobeiri and Westgeest (2024) calls “hypothetical memories,” where AI not only archives but recreates and reinterprets invisible memorial narratives. The generated “synthetic vestiges,” although not direct historical evidence, provide fundamental symbolic and affective value for the construction of a Latin American queer collective memory.

The project was also exhibited at the LGTBI Center of Barcelona as part of the cycle Interferències. memòries col·lectives, which seeks to construct memories away from the single official discourse. As noted in the exhibition documentation, the project presents a retrofuturistic exercise that “uses computational algorithms—technology usually immersed in future narratives—to reimagine our local queer/kuir past and reclaim an archive that could not even exist” (Centro LGTBI de Barcelona, 2024).

This critical-artistic use of AI for the symbolic production of memory represents what Richard (1994) calls a “poetics of crisis,” which allows questioning and subverting hegemonic narratives about the past. The project not only documents a historical absence but, as Erll et al. (2009) suggest, actively “remediates” this absence, creating new layers of meaning and forms of interaction with a systematically erased past.

From the perspective of studies on the intersection between archives and queer theory, Rivas San Martín’s project engages with discussions about the fragility of archives of invisibilized and denied communities, which gives it specific characteristics and mechanisms of enunciation and vindication. Following this reflection, study, and social action, this project can be understood as an “archive of feelings” (Cvetkovich, 2003; Edenheim, 2014). Queer archives are in constant mutation; objects and traces serve as depositories of emotions and affects, with their connection to materials occurring in intimacy before manifesting publicly as evidence of another possible existence. Objects are continuously constructed and deconstructed, allowing affects to reveal themselves through them. This ephemeral condition is what enables archives to be “queered.” Thus, the aim is not only to create collections of queer objects and memories but also to “queer” existing archives (Morris and Rawson, 2013; Zepeda, 2018).

Rivas San Martín’s work demonstrates how AI technology can be appropriated and subverted to serve purposes of historical vindication and collective healing. As a techno-aesthetic apparatus and affective artifact, AI becomes a tool of cultural resistance that allows, as Belk (2013) notes, extending and reconfiguring the collective identity of historically marginalized communities.

7.2 Exhumar la memor.IA (2023)

The project Exhumar la Memor.IA² by Rogelio Séptimo represents a significant case at the intersection between generative AI image technology, collective memory, and affective restoration. This initiative, developed on the island of Janitzio, Michoacán, exemplifies how contemporary techno-aesthetic apparatuses can serve as affective artifacts and tools for reconstructing fractured memory and historical healing of marginalized communities.

The project emerges from a personal motivation of the artist to visualize his great-grandmother, of whom no photographic records existed, and expands toward a community dimension that encompasses the inhabitants of the Lake Pátzcuaro region. As Séptimo himself notes: “In my childhood, family stories made me form an image of her and, in a way, a memory that endures to this day” (MacMasters, 2023). This personal search evolves into a collective memory project that resonates with Hirsch’s (1997) concept of postmemory, where connection with the past is established through imaginative investment and mediated creation.

The project’s methodological strategy incorporates what Assmann (2009) distinguishes as communicative memory and cultural memory. On one hand, it operates through direct family communication via “barter tables,” where participants exchange testimonies and photographs. On the other, it generates cultural artifacts in the form of AI-generated portraits that materialize this collective memory. The technical process combines analog photography with generative AI DALL-E, exemplifying what Van Dijck (2007) calls “mediated memories” in the digital age.

The participatory component of the project, materialized in the barter tables, recovers ancestral exchange practices of the Purépecha community, aligning with what Richard (1994) calls “insubordination of signs.” These practices function as forms of cultural resistance that allow the visibility of marginalized identities and the construction of active memory. As Séptimo explains: “The barter table is a space where the other and I become equal through our past” (Pardo, 2024).

The project demonstrates how generative AI models can function as techno-aesthetic apparatuses in the sense proposed by Déotte (2012), configuring new forms of sensibility and historical perception. The intervention of images with monochromatic marks on glass, using pixels from archival photos, evidence what Didi-Huberman (2012) calls “synthetic vestiges,” which, although not direct historical evidence, provide symbolic and affective value that allows exploring and understanding the past.

It is important to note that Séptimo’s project transcends mere image generation and enters what Shobeiri and Westgeest (2024) calls “hypothetical memories,” where AI not only archives but recreates and reinterprets memorial narratives. This process of visual reconstruction through AI represents a significant shift in the Barthesian conceptualization of the photographic “this has been,” transforming it into a “this could have been” and generative “this is.” As the artist notes, “The tool has served me to give material meaning to the immaterial part of what resides within us” (Pardo, 2024). This materialization of the immaterial through AI suggests new ways of understanding and processing intergenerational trauma, especially in communities that have experienced what Derrida (1997) calls “archontic violence”—the violence exercised on collective memory through institutional control of archives.

Séptimo’s project uniquely exemplifies what Hirsch (2012) calls “the generation of postmemory,” where the transmission of the past is not based on direct remembrance but on imaginative investment and mediated creation. In this case, AI acts as a technological bridge facilitating this intergenerational transmission, allowing ancestors’ experiences and traits to be “remembered” through a complex network of testimonies, family photographs, and generative algorithms. As Sarlo (2005) notes, postmemory acquires particular characteristics in societies that have experienced historical invisibilization. In the case of the Purépecha communities of Lake Pátzcuaro, the absence of photographic records represents not only a technical lack but evidences a form of historical exclusion from visual representation systems.

Furthermore, the project illustrates how AI can function as a tool of “remediation” in the sense proposed by Erll et al. (2009), where new technologies not only preserve but actively transform existing memories. Séptimo’s decision to incorporate these AI-generated images into traditional rituals such as the Day of the Dead demonstrates how emerging technologies can be meaningfully integrated into ancestral cultural practices. As Didi-Huberman (2012) suggests, each image is a vestige that can help revive past experiences, even when historical archives have been destroyed or never existed.

The project’s ethical dimension manifests in its respect for local traditions and transparency in the use of technology. As Belk (2013) notes, digital artifacts can function as extensions of the self that modify our affective experience. In this case, the AI-generated images act as what Piredda (2020) calls “affective artifacts,” facilitating emotional connections with the past and strengthening collective identities.

Séptimo’s work contributes to what Rancière (2009) calls a new “distribution of the sensible,” by democratizing access to visual representation and challenging historical exclusions in image production. The project demonstrates how AI technology can serve humanistic purposes, facilitating the reconstruction of fractured memories and strengthening community bonds.

7.3 IAbuelas (2023)

IAbuelas,³ a project created by publicist Santiago Barros, represents an innovative intersection between the critical-artistic use of AI technology and the creation of historical and collective visual memory in Argentina, specifically about the search for grandchildren who disappeared during the military dictatorship (1976–1983). This artistic-political initiative seeks to reimagine and visualize identities stolen during this violent period in Argentine history, not to replace the scientific identification methods employed by the Grandmothers of Plaza de Mayo, but to create a space for reflection and visibility that contributes to the cause of searching for approximately 300 grandchildren who have not yet been identified (Vallejo, 2023). As Jelin (2016) points out, memory works are fundamental in constructing narratives that allow processing collective social traumas, and they are tools for restoring fractured memory (Didi-Huberman, 2012).

The project’s creative strategy is based on the critical use of Midjourney as an image generation tool, as a techno-aesthetic apparatus (Déotte, 2012; Flusser, 2006, 2011), employing a process of visual fusion between photographs of disappeared parents to generate the current faces of their children, who would today be between 45 and 50 years old. Sebastián Barros based his work on the cases and photographs from the public archive created by the Grandmothers of Plaza de Mayo to help to identify approximately 500 disappeared grandchildren (Criales, 2023). Barros used Midjourney for its capabilities to manipulate age and achieve photorealism, combining historical photographs as references with specific aging prompts that maintained distinctive facial features.

This technological use dialogues with the concept of postmemory (Hirsch, 2012), where new generations process historical trauma through creative and, in this case, digital algorithmic mediations. The technical process involves using public photographs of disappeared parents and implementing specific prompts to generate updated portraits, presented in a format that includes both the generated and original pictures.

The project has generated a significant community response, reaching 12,500 followers on Instagram and establishing a bridge between younger generations and the historical cause of the Grandmothers of Plaza de Mayo. The initiative functions as what Assmann (2008) calls “connective memory,” creating links between individual and collective memory through AI technology and social networks. Following Richard’s (1994) concept of insubordination of signs, IAbuelas represents a form of semiotic insubordination by resignifying the archive of disappeared persons and reappropriating a technology frequently associated with automation and dehumanization for a profoundly human and political purpose. The generation of these faces constitutes an act of resistance against the erasure of identities perpetrated by the military regime, transforming absence into visual presence.

From an ethical perspective, the project maintains a delicate balance by explicitly declaring itself as unofficial, recognizing the limitations of the technology, maintaining links with @abuelasdifusion for consultation cases, using only publicly accessible photographs, and respecting the sensitivity of the topic and the wishes of families. Regarding affective memory, the project constructs what Lévy (1995) calls an “ecology of the virtual,” where technology serves as a medium for expressing collective affects. The AI-generated portraits function as “transitional objects” (Winnicott, 1971) that allow social trauma to be processed and the search for stolen identities to be kept alive.

The initiative transcends mere visual representation to constitute an exercise in “poetic justice” and “artificial imagination,” as described by its creator Santiago Barros, where technology serves as a tool for reconstructing violently severed family bonds (Esto es ¡FA!, 2023). This process of visual reconstitution operates in what Rancière (2009) calls “distribution of the sensible,” altering how we perceive and process the historical memory inherited from the Argentine dictatorship.

IAbuelas ultimately represents a paradigmatic case of how AI technology can be reoriented towards socially significant ends, contributing to processes of historical postmemory and social justice processes. The project provides evidence of the potential that AI technologies have as tools of resistance and memory when used with ethical sensitivity, artistic awareness, and social commitment. This initiative contributes to establishing new paradigms at the intersection of technology, art, community, and collective memory, while also contributing to the visibility of a historical wound pending reparation.

7.4 Synthetic Memories (2023)

Synthetic Memories,⁴ developed by the studio Domestic Data Streamers in Barcelona, represents an innovative intersection between generative artificial intelligence and the preservation of collective memory. This initiative, which began in 2022, has expanded to address the visual reconstruction of memories at risk of disappearing, particularly those affected by forced displacement, political repression, or natural disasters (Domestic Data Streamers, 2023; e-flux Architecture, 2024).

The project’s fundamental objectives are twofold: on one hand, it seeks to preserve undocumented memories through their visual reconstruction and, on the other, to establish intergenerational bridges through the transmission of these personal stories. As its creators note, the project emerges from the need to address the loss of visual memory in contexts of migratory crisis, Alzheimer’s, and collective trauma (García, 2023).

The project’s creative strategy is based on the conceptualization of generative AI models as techno-aesthetic apparatuses (Simondon, 2007; Déotte, 2012). These systems act as mediators that reconfigure the sensitivity and perception of reality. The process involves personalized interviews of 45–60 min, during which prompt engineering is used to translate oral testimonies into images, with an iterative feedback process with participants.

Community involvement is central to the project, manifesting through opening a public office at the Disseny Hub Barcelona (e-flux Architecture, 2024) and collaboration with various vulnerable communities. As Gasulla, co-founder of the project, notes, “the process is very manual, although it has an intense use of technology” (Del Castillo, 2024), emphasizing the importance of the human factor in technological mediation.

From the perspective of ethical construction and humanistic values, the project aligns with what Richard (1994) calls “insubordination of signs,” by facilitating the recovery of fractured identities through the critical use of art and signs. The initiative maintains a strong commitment to informed consent and the dignity of participants, while democratizing access to technology for purposes of historical healing.

Regarding affective memory, the project operates on multiple theoretical levels. Following Didi-Huberman (2012), the generated images act as “synthetic vestiges” which, although not direct evidence, provide symbolic marking with affective and historical value. As Shobeiri and Westgeest (2024) notes, these “hypothetical memories” generated through AI allow formulating hypotheses about complex realities that help negotiate and stabilize deduced visual forms.

The dimension of “punctum” proposed by Barthes (1984) manifests in the ability of the generated images to provoke deep emotional connections. As evidenced in the case of María, an 84-year-old participant, the visualization of her reconstructed memory allowed her to access “a part of her past that she no longer had words to articulate” (García, 2023).

The project contributes to what Rancière (2009) calls a new “distribution of the sensible,” by giving visibility and voice to traditionally marginalized memories. The initiative demonstrates how AI can be used as a techno-aesthetic apparatus that facilitates social transformation and cultural preservation, beyond its merely instrumental function.

The project’s preliminary results, particularly in the context of reminiscence therapy for patients with dementia, suggest a significant positive impact. Eleven out of sixteen patients expressed interest in increasing synthetic memory recreation sessions, although limitations were observed in patients with advanced dementia (García, 2023).

The project’s aesthetic dimension manifests in the hybridization of techniques that evoke photography, drawing, and watercolor, deliberately keeping faces diffuse to avoid confusion with real photographs (García, 2023). This visual ambiguity proves especially effective in evoking memories, as “an unfinished image allows our imagination to complete it, often in line with our memories” (García, 2023). Thus, AI transcends mere technical recreation to become a medium of expression that facilitates emotional connection with memory.

A notable aspect of the project is its application in reminiscence therapy for Alzheimer’s patients, which has attracted the attention of researchers from the Universities of Toronto and British Columbia (Del Castillo, 2024). The process uses specific visual “triggers” that activate affective memory, and it has been discovered that certain ambiguity in the images can be more effective than total clarity. Although these interventions do not stop the progression of the disease, they do “generate spaces of neuronal plasticity” that improve patients’ ability to connect with their environment and personal history (Del Castillo, 2024).

Synthetic Memories represents an innovative application of AI as a techno-aesthetic apparatus and affective apparatus, generating a postmemory that facilitates the reconstruction and preservation of affective memory. Its humanistic approach and ethical commitment position it as a significant model for future initiatives at the intersection of technology, art, and collective memory.

8 Comparative analysis of artistic-technological interventions

The systematic study of the four cases reveals significant patterns in the application of AI for the restoration of affective memory. The analyzed projects show different approaches to the intersection between technology, memory, and affect, evidencing significant convergences and relevant particularities in their conceptualization and implementation.

The objectives and scope of the interventions reflect a spectrum ranging from the individual to the collective. Un archivo queer inexistente addresses the historical absence of representation of a systematically invisibilized community, working at the level of collective memory and historical vindication. IAbuelas operates at the intersection between the familial and the socio-political by visualizing identities stolen by state violence, connecting individual search with collective struggle for memory and justice. Exhumar la Memor.IA is situated in the community-cultural sphere, recovering family memories in a specific cultural context and linking traditional practices with contemporary technologies. Synthetic Memories, for its part, is oriented toward preserving personal memories at risk, with a primarily therapeutic approach that explores the intersection between individual memory and emotional well-being.

The methodological strategies employed reveal different approaches to AI image generation. Un archivo queer inexistente develops a “minority prompt” to counteract algorithmic biases, demonstrating a critical approach to technological limitations. IAbuela implements a process of visual fusion between existing photographs, creating a temporal bridge between historical archive and actuality. Exhumar la Memor.IA stands out for integrating traditional exchange practices with contemporary technology, demonstrating how AI can complement and enrich existing cultural practices. Synthetic Memories employs an iterative feedback process with participants, prioritizing personalization and therapeutic efficacy.

The relationship with affected communities also presents significant variations. Un archivo queer inexistente operates at the level of collective representation, creating a space of visibility for systematically erased histories. IAbuelas maintains a delicate balance between artistic intervention and the institutional work of the Grandmothers of Plaza de Mayo, demonstrating how AI can complement, without replacing, processes of search and reparation. Exhumar la Memor.IA is grounded in pre-existing cultural practices, using technology to strengthen community bonds. Synthetic Memories develops a model of personalized co-creation that prioritizes individual agency in memory reconstruction.

The results and validation mechanisms also differ. Un archivo queer inexistente has found recognition in artistic and activist spaces, validating its critical approach to history and technology. IAbuelas has generated a significant response on social networks and institutional recognition, demonstrating AI’s potential to amplify social causes. Exhumar la Memor.IA is validated through its integration into community cultural practices, while Synthetic Memories documents specific results in therapeutic contexts.

This comparative analysis allows identifying fundamental common elements in the humanistic use of AI: the importance of ethical design, the need for community participation, the search for balance between technological innovation and human sensitivity, and commitment to restoring marginalized or at-risk memories. The differences in approach and results demonstrate AI’s flexibility as a tool for affective memory restoration, adapting to different contexts and needs without losing its transformative potential.

9 Discussion

The analysis of the studied cases reveals significant patterns that illuminate transformative possibilities at the intersection of generative AI, postmemory, and restoration of fractured identities. The findings of this research provide evidence about the use of AI as a tool that enables the construction and reconstruction of lost memories, which function to heal individual or collective traumas. The analyzed cases demonstrate that AI, employed as a techno-aesthetic apparatus (Simondon, 2007; Déotte, 2012) and affective artifact (Belk, 2013; Piredda, 2020), enables a humanistic realization of this technology in service of affective memory in processes of political violence, cognitive deterioration, or social marginalization, as evidenced in the four analyzed cases.

The research reveals that the studied projects show three levels of intervention in fractured memory: the socio-political (Un archivo queer inexistente, IAbuelas), the community-cultural (Exhumar la Memor. IA) and the therapeutic-personal (Synthetic Memories). In all cases, AI transcends its purely instrumental function to constitute itself as a device of symbolic production and postmemory (Hirsch, 1997, 2008). This finding confirms the initial hypothesis about the humanistic potential of AI when employed for memorial and affective restoration purposes.

A central aspect that emerges from the analysis is how these projects operate simultaneously in the register of what Assmann (2009) distinguishes as communicative memory and cultural memory. For example, IAbuelas works in the direct family sphere and in the public sphere of historical archives, while Exhumar la Memor.IA integrates traditional community exchange practices with contemporary technologies. This duality allows the interventions to impact both the preservation of individual memories and the construction of collective narratives.

The analyzed cases demonstrate that technology functions as a medium for the “insubordination of signs” (Richard, 1994), allowing the emergence of alternative narratives that challenge dominant discourses and the “archontic violence” described by Derrida (1997). This political dimension of synthetic memory is particularly evident in Un archivo queer inexistente and IAbuelas, where AI is used to make visible histories systematically erased or denied by hegemonic power structures.

The effectiveness of these projects lies in their ability to generate what Van Dijck (2007) calls “mediated memories” and what Shobeiri and Westgeest (2024) describes as “hypothetical memories.” The results show that AI-generated images, although synthetic, provoke significant affective responses when constructed with sensitivity to specific historical and personal contexts. This ability to generate deep emotional connections is especially manifest in work with communities affected by historical traumas or significant losses, as evidenced by testimonies collected in Synthetic Memories and IAbuelas.

The analysis also reveals the crucial importance of community involvement and ethical design in these projects. The active participation of affected communities legitimizes technological interventions and contributes to what Rancière (2009) calls a new “distribution of the sensible.” This is evident in how projects enable democratic access to visual representation and challenge historical exclusions in image production.

Analyzing the studied cases also reveals critical ethical considerations that must be critically addressed. A central concern is whether AI-generated images could restrict alternative interpretations of the past, particularly considering that memory is a dynamic process of continuous reinterpretation. The risk that synthetic visual representations might crystallize certain narratives at the expense of others deserves special attention, as it could inadvertently generate new forms of “archontic violence” through technological means.

The analyzed cases show various strategies for managing this ethical tension. IAbuelas explicitly acknowledges the non-evidential nature of its portraits, while Un archivo queer inexistente deliberately maintains technical imperfections that reveal the constructed nature of the images. These approaches suggest that the ethical use of AI for memory restoration requires transparency about the constructed nature of the images and their relationship to historical truth.

Likewise, questions arise about the long-term preservation and control of AI-generated memory artifacts. Since most generative AI platforms are controlled by private corporations, it is necessary to consider how generated community memories can be protected from commercial exploitation and ensure their accessibility beyond commercial platforms. This dimension of digital sovereignty constitutes an emerging ethical challenge for AI-based memory projects, especially those oriented toward historically marginalized communities.

These ethical considerations suggest that the humanistic use of AI for memory restoration requires technical skill and a reflective engagement with questions of truth, representation, power, and community agency as these technologies continue to evolve.

In methodological terms, the research demonstrates the effectiveness of combining the analysis of techno-aesthetic apparatuses (Déotte, 2012; Flusser, 2011) with the theory of affective artifacts (Piredda, 2020) to understand how digital objects modify and configure our emotional experience. The studied cases reveal that AI models not only archive or recreate memories but can generate new forms of affective experience that transcend the traditional dichotomy between the real and the synthetic.

The research identifies a significant evolution in the Barthesian conceptualization of the photographic “this has been” towards a generative “this is.” This displacement does not imply a loss of testimonial value, but rather an expansion of the possibilities of memorial construction through what we might call “affective symbolic documentalism.” This new form of documentalism, evidenced in all four analyzed cases, suggests that AI can contribute to the creation of archives that, although not direct historical evidence, possess significant affective and reparative value.

The results also indicate the emergence of new forms of “remediation” (Erll et al., 2009) through AI, where technologies not only preserve but actively transform existing memories. This transformation manifests particularly effectively in projects like Synthetic Memories, where the visual ambiguity of AI-generated images facilitates processes of reminiscence and emotional connection as effective as traditional photographic representations.

The research also reveals the importance of considering what Derrida (1997) calls “archontic violence” in the context of digital memory. The analyzed cases demonstrate that when AI is used in a critical-artistic manner and for humanistic purposes, it can be a tool of resistance against institutional control of memory, allowing the visibility of traditional voices and narratives excluded from official archives. This is a relevant aspect in the Latin American context, where the struggle for historical memory continues to be a field of political and cultural dispute.

Finally, the analysis of these cases suggests that we are facing an emergent phenomenon that we could call “algorithmic postmemory,” where AI actively mediates and participates in the construction of collective and individual memories. This phenomenon poses new ethical and methodological challenges for the study of memory in the digital age, but also offers unprecedented opportunities for the restoration of fractured memories and the healing of historical traumas.

As evidenced throughout the analysis, this research presents significant limitations that must be explicitly acknowledged. In addition to the geographical and temporal delimitation already noted in the methodology, the algorithmic biases inherent to AI systems, discussed in the analysis of Un archivo queer inexistente, represent a significant cultural barrier across the entire field of study. Likewise, the potential risk associated with the creation of artificial memories, although addressed in the ethical considerations, remains a fundamental challenge that requires continued attention in future research.

While the above findings demonstrate AI’s significant potential in restoring affective memory, it is fundamental to recognize and critically analyze the limitations and challenges that these projects face. Examining these restrictions not only contributes to a more complete understanding of the studied phenomenon but also provides key elements for the future development of similar interventions.

The analysis of the case studies reveals the limitations and challenges of AI systems employed for the restoration of affective memory. A crucial aspect is the algorithmic biases inherent to AI models that represent a cultural barrier, particularly in the Latin American context, where training databases may not adequately represent the region’s ethnic and cultural diversity. This technical limitation is especially manifest in projects like Un archivo queer inexistente, where it was necessary to develop specific strategies to counteract the normative tendencies of the models.

The ethical dimension acquires particular relevance in contexts of historical trauma. The potential risk of retraumatization, evidenced in cases like IAbuelas, raises the need to develop careful intervention protocols. The question of informed consent and using personal archives as base material for image generation requires specific ethical considerations, as demonstrated by the experience of Synthetic Memories.

An additional challenge emerges in the tension between “algorithmic truth” and historical memory. AI-generated images, although effective as affective artifacts, cannot and should not replace historical evidence, as explicitly recognized in the IAbuelas project through its creator, Santiago Barros (Esto es ¡FA!, 2023), and its relationship with the Grandmothers of Plaza de Mayo. This tension relates to broader concerns about the long-term preservation of generated digital archives and their future accessibility, considering rapid technological evolution and dependence on commercial platforms.

The analyzed cases demonstrate that when AI is used from a humanistic perspective and with clear ethical commitment, it can contribute significantly to what Hirsch (2012) calls “the work of postmemory,” facilitating the intergenerational transmission of traumatic experiences and the construction of new forms of collective memory. This observation is particularly relevant at a historical moment where the relationship between technology, memory, and affect is profoundly reconfigured by new algorithmic mediation forms.

Analyzing the scope and limitations of these projects allows us to establish significant conclusions about AI’s role in the construction and restoration of affective memory and identify promising lines for future research in this emerging field.

10 Conclusion

This research contributes to the emerging field of AI and affective memory studies by integrating essential theoretical perspectives: postmemory and digital mediation (Hirsch, 1997; Van Dijck, 2007), techno-aesthetic apparatuses (Déotte, 2012; Flusser, 2011), affective artifacts (Belk, 2013; Piredda, 2020), insubordination of signs (Richard, 1994), fractured memory (Didi-Huberman, 2012), punctum as a vehicle of affective memory (Barthes, 1984), and the distribution of the sensible (Rancière, 2009). From this theoretical-methodological integration emerge two original conceptual contributions: the notion of “algorithmic postmemory” as a framework for understanding AI’s active participation in the construction of collective and individual memories, and the concept of “affective symbolic documentalism” that allows understanding the transformation of the testimonial value of the image in the digital age.

The study identifies three levels of intervention in fractured memory through AI: the socio-political, evidenced in projects like Un archivo queer inexistente and IAbuelas; the community-cultural, manifested in Exhumar la Memor. IA; and the therapeutic-personal, developed in Synthetic Memories. This categorization provides a methodological framework for understanding and evaluating future technological interventions in spaces of memory and trauma, especially where traditional narratives have been insufficient or systematically suppressed.

The analyzed cases demonstrate that the effectiveness of these interventions depends on their ability to integrate three fundamental dimensions: ethical sensitivity in the treatment of traumatic memories, commitment to affected communities, and understanding of AI as a techno-aesthetic apparatus capable of generating new forms of affective experience. This triple articulation allows the projects to transcend mere visual recreation to constitute themselves as tools of cultural resistance and collective healing.

However, the research presents essential limitations that must be acknowledged. The selected sample of cases is geographically limited to the Ibero-American context, and although representative, may restrict the generalization of some study findings. The project implementation dates (2022–2023) make it difficult to evaluate their long-term impact. Additionally, the rapid evolution of AI technologies could mean that some specific technical aspects become outdated, although the identified theoretical and methodological principles remain relevant.

This research opens several potential lines for future studies: the development and application of the concept of “algorithmic postmemory” in various cultural contexts; the exploration of “affective symbolic documentalism” as a new form of memorial practice; the study of the relationship between visual ambiguity and therapeutic effectiveness in AI-generated images; and the analysis of ethical and political implications of algorithmic remediation of traumatic memories. Equally relevant would be investigating how these identified levels of intervention could be applied in other contexts of fractured memory.

The research underscores the potential role of AI as a tool that can potentially contribute to social transformation and collective healing when used for humanistic, critical, and artistic purposes. The study’s contribution lies in being able to affirm that the future of affective memory is closely related to the potential offered by AI, not as a replacement for human memory, but as a complement that allows creating new forms of symbolic memorial preservation and restoration. This opens a perspective for the treatment of historical traumas and a critical approach concerning the power of discursive structures associated with controls over memory. Through algorithmic mediation, it is possible to contribute significantly to the construction of more inclusive and reparative memories.

This research opens several potential lines for future studies. The systematic investigation of viewer reactions and community impact would enhance our understanding of reception processes and intergenerational transmission of AI-generated memory artifacts. Expanding applications to cultural contexts beyond the Ibero-American sphere would test the generalizability of the theoretical framework developed here, potentially revealing universal patterns and culturally specific manifestations of algorithmic postmemory across diverse historical and social contexts.

Quantitative assessments of therapeutic efficacy represent another promising research direction, particularly for interventions like Synthetic Memories that operate at the therapeutic-personal level. Such studies could establish evidence-based protocols for the application of these technologies in clinical and community settings. Equally significant would be longitudinal studies tracking the evolution of generated memories over time, addressing crucial questions about the durability and transformation of algorithmic postmemory as technological and social contexts change.

These research directions would contribute to developing a more nuanced understanding of the intersection between technology, memory, and affect while informing ethical and effective practical applications of AI for community healing and historical restoration.

Data availability statement

The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author.

Author contributions

JB: Conceptualization, Formal analysis, Methodology, Project administration, Writing – original draft, Writing – review & editing. DZ: Conceptualization, Formal analysis, Methodology, Writing – original draft. NL: Conceptualization, Methodology, Writing – original draft, Writing – review & editing.

Funding

The author(s) declare that no financial support was received for the research and/or publication of this article.

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Generative AI statement

The author(s) declare that Gen AI was used in the creation of this manuscript. In this research, supervised and filtered use of Claude (Claude 3 Sonnet, Anthropic, 2024) was made as an analytical and editorial support tool. The tool supported comparative analysis between case studies, helping to identify common patterns, differences, and similarities and in the systematic crossing of theoretical concepts with practical cases. AI collaborated in analyzing coherence between text sections and suggested transition paragraphs to improve narrative flow, assisting in academic style and writing revision. All theoretical content, critical analysis, and conceptual development is original to the authors. AI was strictly used as a support tool in analysis, editing, and organization, following ethical and academic guidelines for AI use in research suggested by Penabad-Camacho et al. (2024).

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

1. ^Rivas San Martín’s work can be consulted at these links: https://www.academia.edu/115743256/Un_archivo_inexistente (Rivas San Martín, 2022a, 2022b); https://www.razon.com.mx/cultura/2023/02/13/artista-chileno-usa-inteligencia-artificial-para-crear-obras-como-logro-hacerlo/ (Góchez, 2023).

2. ^The artist’s work can be viewed at https://www.rogelioseptimo.com/exhumar-la-memor-ia/ (Séptimo, 2023).

3. ^The artist’s work can be viewed at https://www.instagram.com/iabuelas/ (Barros, 2023).

4. ^The project can be viewed at https://www.syntheticmemories.net/research/ ((Domestic Data Streamers, 2023).

References

Ajuntament de Barcelona. (2024). “Encarnar la ferida” y “Un archivo inexistente”, dos exposiciones para hacer memoria y dejar registro. Available at: https://ajuntament.barcelona.cat/arxiumunicipal/arxiuhistoric/es/news/encarnar-la-ferida-y-un-archivo-inexistente-dos-exposiciones-para-hacer-memoria-y-dejar-registro-1379550

Google Scholar

Anthropic. (2024). Claude 3 Sonnet [Large language model]. Available at: https://www.anthropic.com/claude

Google Scholar

Assmann, A. (2008). “Canon and archive” in Cultural memory studies: An international and interdisciplinary handbook. eds. A. Erll and A. Nünning (Berlin: De Gruyter), 97–107.

Google Scholar

Assmann, A. (2009). “Memory, individual and collective” in The Oxford handbook of contextual political analysis. eds. R. E. Goodin and C. Tilly (Oxford: Oxford University Press), 210–224.

Google Scholar

Bañuelos, J. (2022). "Evolución del Deepfake: campos semánticos y géneros discursivos 2017-2021" [Evolution of Deepfake: Semantic Fields and Discursive Genres 2017-2021]. Revista ICONO 14 20 (1), 1–27. doi: 10.7195/ri14.v20i1.1771

Crossref Full Text | Google Scholar

Barros, S. (2023). IAbuelas [Instagram profile]. Instagram. Available at: https://www.instagram.com/iabuelas/

Google Scholar

Barthes, R. (1984). Camera lucida: Reflections on photography. Flamingo.

Google Scholar

Belk, R. W. (2013). Extended self in a digital world. J. Consum. Res. 40, 477–500. doi: 10.1086/671052

Crossref Full Text | Google Scholar

Buchloh, B. (1999). “Warburg's paragon? The end of collage and photomontage in postwar Europe” in Deep storage: Collecting, storing and archiving in art (New York: P.S.1), 50–60.

Google Scholar

Centro LGTBI de Barcelona (2024). “Encarnar la ferida” y “Un archivo inexistente”, dos exposiciones para hacer memoria y dejar registro. Ajuntament de Barcelona. Available at: https://ajuntament.barcelona.cat/arxiumunicipal/arxiuhistoric/es/news/encarnar-la-ferida-y-un-archivo-inexistente-dos-exposiciones-para-hacer-memoria-y-dejar-registro-1379550

Google Scholar

Chang, M., Druga, S., Fiannaca, A. J., Vergani, P., Kulkarni, C., Cai, C. J., et al. (2023). “The prompt artists,” in Proceedings of the 15th Conference on Creativity and Cognition. arXiv. p. 75–87. Retrieved March 5, 2025, from https://arxiv.org/abs/2303.12253.

Google Scholar

Criales, J. P. (2023). La inteligencia artificial imagina cómo se verían hoy los bebés robados por la dictadura argentina [Artificial intelligence imagines how babies stolen by the Argentine dictatorship would look today]. El País. Available online at: https://elpais.com/argentina/2023-07-30/la-inteligencia-artificial-imagina-como-se-verian-hoy-los-bebes-robados-por-la-dictadura-argentina.html (Accessed March 5, 2025).

Google Scholar

Cvetkovich, A. (2003). An archive of feelings: Trauma, sexuality, and lesbian public cultures. Durham: Duke University Press.

Google Scholar

Dash, A., and Agres, K. (2024). Ai-based affective music generation systems: A review of methods and challenges. ACM Comput. Surv. 56, 1–34.

Google Scholar

Del Castillo, Carlos (2024). Inteligencia Artificial contra el olvido [Artificial Intelligence Against Forgetting]. elDiario.es. Available online at: https://www.eldiario.es/tecnologia/inteligencia-artificial-olvido_1_2024.html (Accessed May 16, 2024).

Google Scholar

Déotte, J.-L. (2012). ¿Qué es un aparato estético? [What is an Aesthetic Apparatus?]. Santiago: Metales Pesados.

Google Scholar

Derrida, Jacques. Mal de archivo: una impresión freudiana [Archive Fever: A Freudian Impression]. Translated by Vidarte, Paco. Trotta. Madrid. (1997).

Google Scholar

Devillers, L. (2021). “Human-Robot interactions and affective computing: The ethical implications,” in Robotics, AI, and Humanity. eds. J. von Braun, M. S. Archer, G. M. Reichberg, and M. Sánchez Sorondo (Cham: Springer). doi: 10.1007/978-3-030-54173-6_17

Crossref Full Text | Google Scholar

Didi-Huberman, G. (2012). Arde la imagen [Image Burns]. Oaxaca: Ediciones Ve.

Google Scholar

Domestic Data Streamers. (2023). Synthetic memories research. Available at: https://www.syntheticmemories.net/research/

Google Scholar

Drozdzewski, D., and Birdsall, C. (2019). “Advancing memory methods” in Doing memory research: New methods and approaches (Singapore: Palgrave Macmillan), 1–20.

Google Scholar

Edenheim, S. (2014). Lost and never found: the queer archive of feelings and its historical propriety. Differences 24, 36–62. doi: 10.1215/10407391-1892907

Crossref Full Text | Google Scholar

e-flux Architecture. (2024). Citizens’ office of synthetic memories. DHub --Disseny Hub Barcelona. Available at: https://www.e-flux.com/announcements/608285/citizens-office-of-synthetic-memories/

Google Scholar

Erll, A., Ansgar, N., and Ann, R. (2009). Mediation, remediation, and the dynamics of cultural memory. Berlin: De Gruyter.

Google Scholar

Esto es ¡FA! (2023). @IABUELAS, la IA al servicio de la memoria [Video]. YouTube. Available at: https://www.youtube.com/watch?v=mKnuWzx9Wik&t=363s.

Google Scholar

Fernández de Caleya Vázquez, A., and Garrido-Merchán, E. (2024). A taxonomy of the biases of the images created by generative artificial intelligence. arXiv. Available online at: https://arxiv.org/abs/2407.01556 (Accessed March 15, 2025).

Google Scholar

Flusser, Vilém. Towards a philosophy of photography. Translated by Mathews, Anthony. London: Reaktion Books. (2006).

Google Scholar

Flusser, Vilém. Into the universe of technical images. Translated by Roth, Nancy Ann. Minneapolis: University of Minnesota Press. (2011).

Google Scholar

Foucault, Michel. Vigilar y castigar: nacimiento de la prisión [discipline and punish: the birth of the prison]. Translated by Camino, Aurelio Garzóndel. Siglo XXI. México. (1983).

Google Scholar

Freud, S. (1992). “La represión,” in Obras completas. ed. J. Strachey (Amorrortu) 14, 135–153.

Google Scholar

Freud, S. (1980). “El sueño de la inyección de Irma,” in Escritos sobre la cocaína. ed. R. Byck (Barcelona: Anagrama), 254–273.

Google Scholar

García, P. (2023). Capturing the mind's eye: how a Barcelona-based studio is exploring the intersection of memory, art, and technology. Atlantic Dialog. 1, 45–62.

Google Scholar

Garde-Hansen, J. (2011). Media and memory. Edinburgh: Edinburgh University Press.

Google Scholar

Góchez, A. (2023). Artista chileno usa inteligencia artificial para crear obras, ¿cómo logró hacerlo? La Razón.

Google Scholar

Guasch, A. M. (2011). Arte y archivo, 1920–2010: Genealogías, tipologías y discontinuidades. Madrid: Akal.

Google Scholar

Hirsch, M. (1997). Family frames: Photography, narrative, and postmemory. Cambridge: Harvard University Press.

Google Scholar

Hirsch, M. (2008). The generation of postmemory. Poet. Today 29, 103–128. doi: 10.1215/03335372-2007-019

Crossref Full Text | Google Scholar

Hirsch, M. (2012). The generation of postmemory: Writing and visual culture after the holocaust. New York: Columbia University Press.

Google Scholar

Jelin, E. (2016). Las luchas por la memoria. Revista Telar, 17–40.

Google Scholar

Lévy, P. (1995). Qu’est-ce que le virtuel? La Découverte.

Google Scholar

MacMasters, Merry. (2023). "Proyecto con IA gana uno de los premios de 20 Bienal del Centro de la imagen" [AI project wins one of the 20th image center biennial awards]. La Jornada, September 22, 2023.

Google Scholar

Martín Prada, J. (2025). AI-based generative image production systems in the artistic problematisation of the past: the thematisation of memory and temporality in 'AI art'. AI & Soc. doi: 10.1007/s00146-024-02163-z

Crossref Full Text | Google Scholar

Morris, C. E., and Rawson, K. J. (2013). “Queer archives/archival queers” in Theorizing histories of rhetoric. ed. M. Ballif (Carbondale: Southern Illinois University Press), 74–89.

Google Scholar

Pardo, Lucía. (2024). El pasado compartido [The Shared Past]. La Tempestad, October 31, 2024.

Google Scholar

Pataranutaporn, Pat, Danry, Vincent, Blanchard, Lindy, Thakral, Lokesh, Ohsugi, Nozomi, Maes, Pattie, et al. (2023). Living memories: AI-generated characters as digital mementos. In Proceedings of the 28th International Conference on Intelligent User Interfaces, pp. 889–901. New York: ACM Press.

Google Scholar

Penabad-Camacho, L., Morera-Castro, M., and Penabad-Camacho, M. A. (2024). "Guía para uso y reporte de inteligencia artificial en revistas científico-académicas" [Guide for Use and Reporting of Artificial Intelligence in Scientific-Academic Journals]. Rev. Elect. Educ. 28, 1–41. doi: 10.15359/ree.28-1.1

Crossref Full Text | Google Scholar

Piredda, G. (2020). What is an affective artifact? A further development in situated affectivity. Phenomenol. Cogn. Sci. 19, 549–567. doi: 10.1007/s11097-019-09628-3

Crossref Full Text | Google Scholar

Rancière, Jacques. (2009). El reparto de lo sensible: Estética y política [the distribution of the sensible: Aesthetics and politics]. Translated by Padró, Miguel. Santiago: LOM ediciones.

Google Scholar

Richard, N. (1994). La insubordinación de los signos: cambio político, transformaciones culturales y poéticas de la crisis [The Insubordination of Signs: Political Change, Cultural Transformations, and Poetics of Crisis]. Santiago: Editorial Cuarto Propio.

Google Scholar

Richard, N. (2011). The political in art: art, politics and institutions. e-misférica 6, 1–16.

Google Scholar

Richard, N. (2013). Fracturas de la memoria: Arte y pensamiento crítico [Fractures of Memory: Art and Critical Thinking]. Buenos Aires: Siglo XXI.

Google Scholar

Rivas San Martín, F. (2022a). "Un archivo queer inexistente" [A Nonexistent Queer Archive]. Available online at: Academia.edu and https://academia.edu/documents/queer-archive (Accessed January 10, 2025).

Google Scholar

Rivas San Martín, F. (2022b). Tecnopoéticas del activismo y la disidencia [Technopoetics of activism and dissidence]. Santiago: La Escuela.art.

Google Scholar

Sarlo, B. (2005). Tiempo pasado: Cultura de la memoria y giro subjetivo [Past Time: Memory Culture and Subjective Turn]. Buenos Aires: Siglo XXI.

Google Scholar

Sekula, A. (1986). The body and the archive. JSTOR 39, 3–64. doi: 10.2307/778312

Crossref Full Text | Google Scholar

Séptimo, R. (2023). Exhumar la Memor.IA. Available online at: https://www.rogelioseptimo.com/exhumar-la-memor-ia/ (Accessed March 2, 2025).

Google Scholar

Shobeiri, A., and Westgeest, H. (2024). Virtual photography: Artificial intelligence, in-game, and extended reality. Bielefeld: transcript Verlag.

Google Scholar

Simondon, G. (2007). El modo de existencia de los objetos técnicos [the mode of existence of technical objects]. Translated by Margarita Martínez. Buenos Aires: Prometeo.

Google Scholar

Vallejo, Mariana (2023). IAbuelas: cómo imaginar el rostro de un desaparecido [IAbuelas: how to imagine the face of a disappeared person]. Vist projects, Available online at: https://vistprojects.com/iabuelas-2023 (Accessed July 31, 2023).

Google Scholar

Van Dijck, J. (2007). Mediated memories in the digital age. Stanford: Stanford University Press.

Google Scholar

Winnicott, D. W. (1971). Playing and reality. London: Routledge.

Google Scholar

Young, J. E. (2000). At memory's edge: After-images of the holocaust in contemporary art and architecture. New Haven: Yale University Press.

Google Scholar

Zepeda, L. (2018). Queering the archive: transforming the archival process. DisClosure 27, 17–30. doi: 10.13023/disclosure.27.03

Crossref Full Text | Google Scholar