Your new experience awaits. Try the new design now and help us make it even better

ORIGINAL RESEARCH article

Front. Hum. Dyn., 05 February 2026

Sec. Digital Impacts

Volume 8 - 2026 | https://doi.org/10.3389/fhumd.2026.1695355

This article is part of the Research TopicAffective and Behavioral Dynamics in Human-Technology Interactions of Industry 5.0View all 12 articles

How AI usage shapes employees’ exploitative and exploratory innovation: joy as a mediator and learning goal orientation as a moderator

Li Liu
Li Liu1*Ya LvYa Lv2
  • 1Dhurakij Pundit University, Bangkok, Thailand
  • 2Credit Approval Department, Fujian Branch, Postal Savings Bank of China, Fuzhou, Fujian, China

Introduction: Employees play a pivotal role in organizational ambidextrous innovation, yet existing studies have paid limited attention to how artificial intelligence shapes employees’ exploitative and exploratory innovation. Drawing on cognitive appraisal theory and the broaden-and-build theory of positive emotions, this study identifies joy as the central emotional mechanism linking artificial intelligence usage to these two forms of innovation. As a high-arousal positive emotion grounded in person–situation fit, joy promotes active engagement, in contrast to lower-arousal emotions such as satisfaction or happiness that reflect acceptance rather than pursuit. We further examine how learning goal orientation moderates the extent to which joy translates into exploitative and exploratory innovation, thereby advancing understanding of how technological empowerment affects ambidextrous innovation.

Methods: This study draws on survey data from Chinese employees (N = 669) and employs partial least squares structural equation modeling (PLS-SEM) to examine the mediating role of joy in the relationship between artificial intelligence usage and employees’ exploitative and exploratory innovation, as well as the moderating effect of learning goal orientation.

Results: Artificial intelligence usage shows small but meaningful positive effects on employee exploitative innovation (β = 0.120) and exploratory innovation (β = 0.104), with joy partially mediating the effect on exploratory innovation only (indirect β = 0.050). Moreover, joy positively predicts exploitative innovation (β = 0.182) and exploratory innovation (β = 0.206) only under high learning goal orientation.

Discussion: The findings emphasize that the role of positive emotions is not universal but rather motivation-dependent: joy mediates exploratory but not exploitative innovation, while high learning goal orientation amplifies its effects on both innovation types. These results extend emotion theories to AI-enabled work contexts and offer practical implications for fostering employees’ emotions and learning motivation to achieve synergy between technological empowerment and innovation.

1 Introduction

In the context of accelerated digital and intelligent transformation, enterprises aiming for sustainable development in a dynamic competitive environment must simultaneously deepen existing capabilities (exploitative innovation) and actively explore new opportunities (exploratory innovation; Lauenstein et al., 2025). However, these two types of innovation often compete for limited resources, making it difficult to balance them effectively (Denrell et al., 2025). Prior research indicates that employees are central to both exploitative and exploratory innovation, and can foster organizational ambidextrous innovation from the bottom up by flexibly switching between the two (Boemelburg et al., 2023; Mom et al., 2019; Otto et al., 2024).

With artificial intelligence (AI) widely embedded in organizational business processes, employees’ working patterns, ways of acquiring information, and role perceptions in innovation activities are undergoing profound changes (Ocal and Crowston, 2024). The usage of AI exhibits a typical “double-edged sword” effect: on the one hand, it can significantly enhance work efficiency and satisfaction, providing cognitive and resource support for employee innovation (Eshraghian et al., 2024; Noy and Zhang, 2023); on the other hand, empirical research has shown that overreliance on AI may weaken employees’ autonomous thinking and deep information processing ability, thereby suppressing their innovation potential (Burton et al., 2024; Kanbach et al., 2023). Taken together, these mixed effects suggest that the consequences of AI usage cannot be fully understood solely in terms of its technical features, but also depend on how employees cognitively perceive and emotionally experience AI in their work context. Accordingly, a critical research question concerns whether and how AI usage promotes employees’ exploitative innovation (EEI) and exploratory innovation (EXI).

Existing studies on ambidextrous innovation have predominantly focused on traditional resources, such as organizational support, leadership styles, and individual characteristics (Boemelburg et al., 2023; Hardy et al., 2024; Otto et al., 2024; Zhang et al., 2023). However, how AI, as an emerging digital resource, systematically shapes employees’ ambidextrous innovation remains underexplored, particularly regarding the differentiated pathways through which it affects EEI and EXI. To address these gaps, drawing on prior research highlighting employees’ technological perceptions and emotional responses to AI usage (Noy and Zhang, 2023), this study examines how employees’ cognitive appraisals and affective processes contribute to the distinct pathways through which AI usage influences EEI and EXI.

Emotions, as critical outcomes of cognitive appraisal, are recognized to influence innovative performance through their impact on information processing (Johnson, 2020). Prior studies indicate that employees tend to hold a positive attitude toward AI usage (Guha et al., 2023; Eshraghian et al., 2024). While such positive attitudes are often associated with low-arousal positive states (e.g., satisfaction and happiness) reflecting efficiency and convenience, AI usage can also elicit high-arousal positive emotions (e.g., joy) when employees cognitively appraise it as supporting competence development and meaningful goal achievement. Accordingly, focusing on discrete positive emotions offers a theoretically informative perspective for understanding how AI usage shapes employees’ ambidextrous innovation.

The broaden-and-build theory of positive emotions (BBT) proposes that high-arousal positive emotions expand individuals’ cognitive structures, facilitating information integration and exploratory tendencies (Fredrickson, 2013). Cognitive appraisal theory (CAT) further underscores that various positive emotions are linked to specific appraisal patterns, which in turn elicit distinct coping tendencies (Lazarus, 1991). Among these, joy, as a high-arousal positive emotion, typically arises from a sense of accomplishment derived from the alignment between technology usage and personal goals. It broadens cognitive and behavioral repertoires, while also reinforcing psychological resilience and sustaining long-term innovative motivation (Johnson, 2020; Tan and Titova, 2024). In addition, the influence of emotions is contingent upon individual motivational traits (Smith and Lazarus, 1993), learning goal orientation (LGO), characterized by an emphasis on competence development and task mastery, may reinforce the extent to which joy facilitates both EEI and EXI (Marshall et al., 2019).

In summary, drawing on CAT and the BBT, this study develops a moderated mediation model of “AI usage - joy - EEI/EXI.” Accordingly, this study pursues two specific research objectives: (1) to examine whether and how joy mediates the relationship between AI usage and EEI/EXI, and (2) to investigate whether LGO moderates the effects of joy on EEI and EXI. By elucidating the synergistic mechanism between positive emotions and motivational dispositions, this study contributes to advancing theoretical insights into human–machine collaboration and ambidextrous innovation, and provides practical implications for enterprises in designing targeted strategies for AI deployment and human resource management to achieve a dynamic balance between technological empowerment and innovation outcomes.

2 Literature review and hypothesis development

AI refers to technologies that mimic human intelligence by replicating cognitive functions to accomplish complex tasks (Davenport et al., 2020). AI usage refers to the extent to which employees integrate AI systems or tools into their work processes, thereby facilitating information processing, knowledge creation, and decision-making (Man Tang et al., 2022).

Existing research generally affirms the value of AI in enhancing organizational efficiency and stimulating innovation; however, the mechanisms by which AI shapes employee-level innovation remain contested (Liu et al., 2024; Mithas et al., 2022). Importantly, organizational ambidexterity arises from balancing exploitative and exploratory innovation, yet the resource competition between these two activities makes their simultaneous pursuit inherently challenging (Mom et al., 2019). Although structural separation across organizational units or temporal sequencing has been proposed as a solution, such arrangements are often difficult to implement in resource-constrained small and medium-sized enterprises (Boemelburg et al., 2023). Consequently, an increasing body of research emphasizes the critical role of employees in enacting ambidexterity from the bottom up. By exercising autonomous judgment in their work, employees flexibly allocate attention and effort between refining existing practices and experimenting with new approaches, which constitutes the microfoundations of organizational ambidextrous innovation (Otto et al., 2024). Accordingly, understanding how AI usage shapes EEI and EXI becomes critical for explaining how digital technologies contribute to organizational ambidexterity.

While concerns persist that excessive reliance on AI usage may dampen employees’ willingness and capacity to innovate (Burton et al., 2024), growing evidence highlights its benefits in enhancing knowledge sharing, autonomy, and information processing (Jia et al., 2023; Malik et al., 2021). Notably, these existing studies adopt a unidimensional view of innovation, overlooking the distinct cognitive and resource demands of EEI and EXI. As a result, the mechanisms by which AI usage differentially influences the two forms of innovation remain largely underexplored (Hwang and Wu, 2025). Given the fundamental distinction between process optimization and disruptive breakthroughs, treating innovation as unidimensional fails to capture AI’s functional role across diverse forms of innovation (Boemelburg et al., 2023; Mom et al., 2019).

Positive emotions have been shown to enhance individuals’ cognitive flexibility and creative potential, thereby providing essential psychological resources for innovation (Fredrickson, 2013). However, positive emotions differ in their arousal level and functional implications (Keltner et al., 2019). Low-arousal positive states such as happiness or satisfaction tend to signal comfort and acceptance of current conditions, whereas joy, as a high-arousal positive emotion, reflects an activated appraisal of person-situation fit that mobilizes approach-oriented motivation and innovation-oriented effort (Arnett, 2023; Tan and Titova, 2024). Accordingly, joy is particularly relevant for understanding different forms of innovation. AI usage can evoke joy by alleviating cognitive load, enhancing perceived control, and improving efficiency, which in turn fosters EEI and EXI (Eshraghian et al., 2024). However, research has largely focused on positive emotions in general, paying little attention to how joy—as a discrete emotion—differentially influences EEI and EXI (Eshraghian et al., 2024; Lin and Chen, 2024). For example, joy may foster EEI through efficiency and task engagement, while facilitating EXI through cognitive flexibility, resource accumulation, and psychological resilience.

2.1 Theoretical background

According to CAT, positive emotions originate from individuals’ primary appraisals of a situation as favorable to their goal attainment and are subsequently refined into distinct types of positive emotions through secondary appraisals, such as responsibility attribution, coping potential, and future expectations (Lazarus, 1991). Among them, joy is characterized as a high-arousal state driven by goal congruence and a sense of accomplishment, manifested through excitement and vitality, with its core rooted in the perceived fit between the individual and the situation (Johnson, 2020). In contrast to happiness (enduring and stable), satisfaction (low-arousal fulfillment), gratitude (externally triggered), or schadenfreude (marked by moral undertones), joy is immediate, motivation-driven, and socially meaningful (Arnett, 2023). Accordingly, joy is not merely a positive affective state but also a dynamic interaction between individuals and their environment, which motivates the pursuit of further goal-congruent experiences and thus demonstrates unique research value in the context of innovation (Tan and Titova, 2024).

The BBT delineates two core functions of positive emotions (Fredrickson, 2001). First, positive emotions broaden individuals’ momentary attentional and cognitive scope, stimulating a wider range of thought patterns and action possibilities and thereby fostering flexible cognition and the capacity to integrate diverse information and materials. Second, positive emotions build enduring personal resources that span psychological, cognitive, and social domains and persist beyond the emotional episode, thereby strengthening resilience and supporting long-term adaptive functioning, including sustained innovative potential (Fredrickson, 2013; Lin and Chen, 2024).

The integration of these two theories provides a systematic account of how AI usage promotes EEI and EXI through joy. Specifically, CAT explains that employees are more likely to experience joy when they cognitively appraise AI usage as reshaping task characteristics by enhancing task controllability and feedback clarity, thereby increasing goal congruence and perceived controllability. Building on this emotional response, the BBT clarifies that joy facilitates innovation through two complementary functions. First, joy broadens attentional and cognitive scope, enhancing flexible information processing and integration, which supports the efficiency-oriented refinement and improvement characteristic of EEI. Second, joy builds enduring personal resources that strengthen resilience and sustain adaptive functioning over time, thereby enabling continued engagement with EXI. Moreover, LGO, as a motivational trait, may amplify the positive effect of joy on EEI and EXI, underscoring individual motivation as a critical boundary condition in the emotion–innovation pathway.

2.2 AI usage and employees’ exploitative innovation

EEI primarily relies on existing knowledge and resources to improve established capabilities or processes in order to enhance efficiency. Compared with exploratory innovation, it entails relatively lower difficulty and risk and requires less complex knowledge structures (Boemelburg et al., 2023; Mom et al., 2019). AI usage plays a pivotal role in this process by streamlining routine and programmed tasks, thereby freeing employees’ time and energy to focus on core activities and strengthening their capacity for knowledge-based improvements (Zhou et al., 2023). Moreover, by simulating human cognitive functions, AI helps employees identify suboptimal processes and propose improvements, reducing cognitive load and enhancing the efficiency of process optimization and product refinement (Dennis et al., 2023). In addition, AI’s ability to store and integrate vast amounts of data facilitates the recombination of existing knowledge, ultimately boosting both productivity and creativity (Lee and Chung, 2024; Peres et al., 2023).

Drawing on CAT and the BBT, employees are likely to appraise the functional benefits of AI usage as enhancing task controllability and goal attainment, which broadens their momentary cognitive scope and supports the efficient refinement of existing knowledge and routines, thereby creating favorable psychological and cognitive foundations for exploitative innovation. Accordingly, we propose the following hypothesis:

H1: AI usage positively influences EEI.

2.3 AI usage and employees’ exploratory innovation

EXI centers on the creation of new knowledge, opportunities, and technologies. It is inherently uncertain, difficult, and risky, requiring employees to mobilize stronger intrinsic motivation and more diverse knowledge structures. Despite these challenges, it delivers substantial benefits and supports the achievement of long-term organizational goals (Boemelburg et al., 2023; Mom et al., 2019). AI usage, with its vast data pools, cross-domain knowledge bases, and advanced algorithms, contributes to EXI in several parallel ways. First, it equips employees with novel analytical insights, extends professional expertise, and unlocks creativity and decision-making capacity (Kanbach et al., 2023; Spring et al., 2022). Second,it enhances motivation and confidence to pursue new solutions by automating routine tasks and supporting complex activities (Malik et al., 2023). Third, AI applications in personalized knowledge management and digital learning strengthen employees’ skills and engagement, thereby fostering innovative thinking (Jatobá et al., 2023; Verma and Singh, 2022). Finally, AI usage stimulates employees’ proactive engagement by enabling high-complexity and specialized tasks that create challenging environments, while its data-driven decision-making and information-processing capabilities facilitate the emergence of new technologies and products (Jia et al., 2023).

Integrating CAT and the BBT, employees are likely to appraise AI usage in uncertain and complex tasks as enhancing coping potential, competence development, and future opportunities, which broadens their cognitive and attentional scope and facilitates the accumulation of enduring cognitive and psychological resources, thereby creating favorable conditions for sustained engagement in exploratory innovation. Accordingly, we propose the following hypothesis:

H2: AI usage positively influences EXI.

2.4 Joy as a mediator

Joy is a high-arousal positive emotion that evokes experiences of ease and freedom, broadens attention and thought processes, and thereby enhances creativity and cognitive flexibility (Johnson, 2020). According to CAT, joy emerges from individuals’ holistic evaluations of goal congruence, responsibility attribution, coping potential, and future expectations within a given situation (Lazarus, 1991). When employees perceive that AI usage enhances efficiency, facilitates knowledge acquisition, and improves performance, and attribute this achievement to their effective mastery of AI, joy is more likely to be evoked (Luo et al., 2021; Shao et al., 2024). Moreover, the continuous advancement of AI’s analytical and predictive capabilities reinforces employees’ positive expectations of its future value, thereby amplifying the experience of joy.

The BBT further elucidates how joy exerts differentiated effects on distinct forms of innovation through its emotional functions. With respect to EEI, joy broadens individuals’ momentary thought–action repertoires, providing immediate cognitive benefits (Fredrickson, 2001, 2013). This momentary broadening facilitates the integration of diverse information into existing knowledge frameworks and supports efficiency-oriented refinement and optimization of processes and products (Forgas and George, 2001; King, 2020). However, because EEI largely relies on established routines, procedural knowledge, and incremental improvements aimed at short-term efficiency gains (Boemelburg et al., 2023), its implementation is less dependent on sustained emotional arousal.

In contrast, EXI requires employees to move beyond existing frameworks and pursue novel directions, involving sustained resource investment and exposure to high levels of uncertainty and risk (Boemelburg et al., 2023). Although joy is inherently transient, its broadening effects promote open and flexible thinking and facilitate the accumulation of psychological resources, thereby enhancing resilience, risk tolerance, and sustained engagement under uncertain conditions (Fredrickson, 2001, 2013). Through these processes, joy supports employees in deriving positive meaning from setbacks and persisting in the pursuit of long-term goals such as exploratory innovation (Johnson, 2020; Tan and Titova, 2024; Welpe et al., 2012).

Taken together, joy functions as a positive emotional mechanism linking AI usage to employee innovation, but its functional significance differs across innovation types. Whereas joy provides supplementary cognitive support for EEI, it constitutes a more fundamental psychological resource for sustaining long-term engagement in EXI. Accordingly, we propose the following hypothesis:

H3: Joy mediates the relationship between AI usage and EEI.

H4: Joy mediates the relationship between AI usage and EXI.

2.5 Learning goal orientation as a moderator

LGO reflects an individual’s inclination to enhance personal competence and to understand or master new tasks. As a key motivational construct, it fosters learning, knowledge acquisition, and creativity (Dweck and Leggett, 1988). From the perspective of CAT, emotional responses arise from individuals’ subjective evaluations of situational events, and this appraisal process is shaped not only by external task characteristics but also by internal motivational orientations, goals, and cognitive capacities (Schwarz, 2012; Smith and Lazarus, 1993). Accordingly, individual differences such as LGO play a central role in shaping how employees interpret AI-enabled work contexts and the meaning they assign to emotional experiences such as joy. Compared with alternative motivational orientations that emphasize outcome validation or risk avoidance, LGO is especially well suited to AI-enabled work because it supports autonomous judgment regarding when to refine existing practices and when to pursue novel solutions, thereby strengthening the translation of joy into both exploitative and exploratory innovation (Boemelburg et al., 2023).

In particular, employees with high LGO prioritize capability development, willingly embrace challenging goals, and exhibit strong motivation and autonomy (Marshall et al., 2019). From the BBT, such individuals are more likely to use joy elicited in AI-enabled work to broaden their momentary cognitive scope and actively engage with task demands, thereby strengthening perceived task control and deepening involvement in process optimization and improvement, which facilitates exploitative innovation (Da Motta Veiga and Turban, 2014). Furthermore, high LGO individuals tend to engage in deep processing and continuous learning, which allows them to convert the broadening effects of joy into enduring cognitive and psychological resources, such as persistence, confidence, and resilience. These built resources sustain engagement with complex and unfamiliar tasks, fostering perseverance and creativity in exploratory innovation (Da Motta Veiga and Turban, 2014; To et al., 2015). By contrast, employees with low LGO are inclined to settle for the status quo, relying on the convenience of AI to complete basic tasks and avoid risks. Without a strong mastery-oriented motivation, joy is less likely to be transformed into broadened cognition or accumulated resources, thereby constraining its positive influence on both exploitative and exploratory innovation (Lench et al., 2016).

Taken together, learning goal orientation shapes how employees interpret and utilize joyful experiences elicited by AI usage, thereby influencing the extent to which joy can be translated into exploitative and exploratory innovation. Accordingly, we propose the following hypothesis:

H5: LGO positively moderates the relationship between joy and EEI, such that the effect of joy on EEI is stronger when LGO is higher.

H6: LGO positively moderates the relationship between joy and EXI, such that the effect of joy on EXI is stronger when LGO is higher.

The research model is depicted in Figure 1.

Figure 1
Flowchart illustrating the relationships between AI usage, joy, learning goal orientation, and employees' innovation types. AI usage positively influences joy (H3) and employees' exploratory (H2) and exploitative (H1) innovation. Joy has a mediation effect on both innovation types (H3, H4). Learning goal orientation positively impacts employees' exploitative (H5) and exploratory innovation (H6). Arrows indicate directionality and hypotheses are marked with H1 through H6.

Figure 1. Hypothesized research model.

3 Methods

3.1 Data collection

To examine the proposed hypotheses, this study employed a quantitative research design, distributing questionnaires both online and offline. Data were collected between January and May 2025 from full-time employees who used AI tools in their daily work.

To enhance methodological rigor, a clear definition of AI usage was provided at the beginning of the questionnaire. AI was defined as digital systems or tools used by employees to support their work, including content generation, problem solving, decision support, or innovation-related tasks, and may involve technologies such as machine learning, natural language processing, image recognition, knowledge-based systems, robotic process automation, or generative AI.

Respondents were then presented with a screening item indicating whether they primarily used AI to support problem solving, task automation, content generation, or other innovation-related activities, or whether they mainly relied on other software tools (e.g., Excel, Photoshop, or traditional information systems) to support their work. Only respondents who selected the former option were allowed to proceed to the main questionnaire, whereas those selecting the latter option were automatically screened out. This screening item was used solely for sample selection purposes and was not treated as a latent construct in the analytical model.

Notably, this study did not differentiate between specific AI types, as the focus was on employees’ functional use of AI as a work-enabling resource rather than on technical distinctions among AI systems. Respondents also indicated their daily AI usage time using predefined time categories, which was used to describe AI usage intensity in the sample.

The survey was conducted in Suzhou and Shanghai, regions in China where AI adoption is relatively advanced. Target enterprises were identified through company websites to ensure that participating organizations had implemented AI in work processes. After obtaining informed consent, questionnaire links were distributed to employees. The incentives consisted of small gifts valued at approximately RMB 30 and were provided solely to encourage participation and voluntary referrals, independent of respondents’ answers or survey completion. Given that the survey was anonymous and posed minimal risk, formal ethical approval was exempted under the institutional guidelines. The study was conducted in accordance with generally accepted ethical principles for research involving human participants.

Following established practices for detecting careless responses in survey research (Meade and Craig, 2012), multiple data screening criteria were applied. Responses with completion times of less than 2 min were excluded, whereas responses exceeding 15 min were treated as missing. Cases showing largely uniform response patterns or clear cross-item inconsistencies between reported daily AI usage time categories and AI usage frequency were removed. After screening, 669 valid responses were retained for further analysis, yielding an effective response rate of 87.64%.

Table 1 presents the sample characteristics, indicating that the respondents’ demographic attributes are broadly representative of the target population. Of the 669 respondents, the gender distribution was nearly balanced (51.6% male; 48.4% female). The majority were born after 2000 (46.5%) or after 1990 (45.0%), reflecting the age profile typical of emerging technology users. Most participants held a bachelor’s degree (61.4%), had 2–5 years of work experience (65.6%), and were employed in IT (28.8%), finance (21.8%), or film/media (19.6%). Frontline employees comprised the majority (81.6%), with relatively few in managerial positions. AI usage intensity was high, with 49.0% using AI 4–8 h daily and 25.4% using it more than 8 h.

Table 1
www.frontiersin.org

Table 1. Participant characteristics.

Although respondents may be nested within organizations, the present study focuses on individual-level cognitive and emotional processes. All focal constructs were conceptualized and measured at the individual level; accordingly, following established practices in management research (Hitt et al., 2007), single-level analysis was adopted.

3.2 Statistical procedure

Partial least squares structural equation modeling (PLS-SEM) was employed to test the proposed model. This approach was appropriate given the presence of multiple latent constructs, mediating and moderating relationships, and a complex path structure, as well as its robustness to potential deviations from multivariate normality (Hair et al., 2019).

Following established PLS-SEM guidelines, we adopted a two-stage analytical procedure. The measurement model was first evaluated in terms of indicator reliability, internal consistency reliability, convergent validity, and discriminant validity to ensure that the constructs were measured appropriately before interpreting structural relationships. Subsequently, the structural model was assessed to test the hypothesized direct, indirect, and moderating relationships among constructs. Given that PLS-SEM does not rely on distributional assumptions, the significance of these effects was examined using a non-parametric bootstrapping procedure with 5,000 resamples, which provides empirical standard errors and confidence intervals for parameter estimates, with p-values reported as complementary information to facilitate interpretation.

In addition, supplementary robustness analyses were conducted to assess the stability of the results across alternative analytical specifications. These analyses are reported in a later section of the manuscript and are intended to examine whether the main findings are sensitive to model assumptions.

3.3 Measurement

The measurement instruments for this study were developed from well-established scales that have been widely adopted in leading journals and repeatedly validated, ensuring strong reliability and validity. To align with the research context, the measures for AI usage, EEI, EXI, joy, and LGO were carefully adapted. All items were translated using a back-translation procedure to ensure semantic equivalence and were reviewed by academic experts to enhance clarity and contextual relevance. In addition, a pilot test was conducted to assess item comprehensibility and refine wording where necessary. Each construct was measured using a five-point Likert scale (1 = strongly disagree, 5 = strongly agree).

In line with the data collection procedure, only respondents who reported active use of AI tools in their work were included in the analysis. The screening item used to identify eligible respondents was employed solely for sample selection and was not specified as a construct in the analytical model.

AI usage: This construct was measured with a three-item scale adapted from Man Tang et al. (2022). A sample item is, “I use artificial intelligence to accomplish most of my work tasks” (Cronbach’s α = 0.803).

Joy: This construct was measured with a seven-item unidimensional scale adapted from Sun et al. (2022), which was originally derived from the emotion dimension scale of Clark and Watson (1994). Despite some linguistic overlap with general positive affect (e.g., happiness), the inclusion of activation-related items (e.g., excited, enthusiastic, lively, energetic) captures high-arousal joy. Thus, joy reflects an activated emotional state elicited by AI usage in this study (Cronbach’s α = 0.890).

Employees’ exploitative and exploratory innovation: EEI and EXI represent the two key dimensions of ambidextrous innovation as conceptualized by Mom et al. (2019). EEI was measured with seven items adapted from Mom et al. (2019); a sample item is, “I have optimized existing processes based on my prior work experience” (Cronbach’s α = 0.887). EXI was measured with seven items from the same source; a sample item is, “I search for possibilities of new services, products, processes, or markets” (Cronbach’s α = 0.888).

Learning goal orientation: This construct was measured with a six-item scale developed by VandeWalle (1997). A sample item is, “I often read materials related to my work to improve my competence” (Cronbach’s α = 0.887).

Control variables: To reduce potential confounding effects and enhance the robustness of model estimation, this study controlled for variables commonly included in prior research on AI usage and employee innovation, following Man Tang et al. (2022) and Bernerth and Aguinis (2016), including gender, age, education level, and work experience. In addition, industry type, job position, and daily AI usage time were controlled to account for potential contextual influences. All control variables were measured using single-item indicators and incorporated as exogenous control variables in the analytical model.

3.4 Common method bias

Given that the data were collected using a cross-sectional, self-report survey, common method bias (CMB) may represent a potential concern. Following the recommendations of Podsakoff et al. (2003), this study addressed potential common method bias through both procedural remedies and an additional statistical diagnostic.

With respect to procedural remedies, anonymity and confidentiality were emphasized in the questionnaire instructions, and respondents were informed that there were no right or wrong answers to reduce evaluation apprehension and social desirability bias. In addition, survey items were refined through expert review and pilot testing, and items measuring different constructs were presented in a mixed order rather than grouped by construct to reduce respondents’ ability to infer hypothesized relationships.

As a statistical diagnostic, Harman’s single-factor test was conducted. The results indicated that no single factor accounted for the majority of the variance, with the first factor explaining 32.51% of the total variance, which is below commonly accepted thresholds. These findings suggest that common method bias is unlikely to substantially bias the results.

4 Results

4.1 Reliability and validity test

This study employed SmartPLS 4 and Mplus 8 for data analysis. As shown in Tables 2, 3, all constructs exhibited acceptable reliability and validity (Cronbach’s α > 0.80, CR > 0.70, AVE > 0.50). Confirmatory Factor Analysis (CFA) results indicated a good fit for the five-factor model (χ2/df = 1.050, CFI = 0.998, TLI = 0.998, RMSEA = 0.009, SRMR = 0.026), supporting discriminant validity (Williams et al., 2010).

Table 2
www.frontiersin.org

Table 2. Measurement model results.

Table 3
www.frontiersin.org

Table 3. Confirmatory factor analysis results for discriminant validity.

4.2 Research model test

The results of the partial least squares structural equation modeling (PLS-SEM) analysis are presented in Table 4. Overall, the findings provide partial support for the proposed hypotheses.

Table 4
www.frontiersin.org

Table 4. Results of the hypothesis tests.

Regarding the main effects, AI usage exerted a significant positive impact on both EEI (H1: β = 0.120, p < 0.01) and EXI (H2: β = 0.104, p < 0.05). Although these effects are modest in magnitude, they indicate that AI usage contributes incrementally to both forms of innovation, explaining approximately 11.4% of the variance in EEI and 11.6% in EXI.

With respect to the mediating role of joy, the indirect effect of AI usage on EXI was statistically significant (H4: β = 0.050, p < 0.05). For EEI, although the bootstrap confidence interval suggested statistical significance, the effect size was small and the p value was marginal (H3: β = 0.041, p > 0.05). Consistent with the recommendations of Hair et al. (2019), this indirect effect was therefore interpreted as statistically weak and lacking substantive relevance, and was not regarded as evidence of a supported mediation.

Regarding the moderating effect, results showed that the interaction between joy and LGO was significant for both EEI (H5: β = 0.089, p < 0.05) and EXI (H6: β = 0.092, p < 0.05). The subsequent simple slope analyses showed that joy promoted both forms of innovation at high levels of LGO (EEI: β = 0.182, p < 0.01; EXI: β = 0.206, p < 0.01), but not at low levels (EEI: β = 0.004, p > 0.05; EXI: β = 0.022, p > 0.05).

For a more intuitive demonstration of the moderating effect of LGO, this study employed simple slope analysis, plotting graphs at the mean (M), −1 SD, and +1 SD to represent average, low, and high values of the variable. The graphs illustrate how employees with high versus low LGO differ in exploitative and exploratory innovation (see Figures 2, 3).

Figure 2
Line graph titled “LGO x JOY” showing the relationship between JOY (x-axis) and EXI (y-axis). Three lines represent LGO at different levels: -1 SD (dashed), mean (dotted), and +1 SD (solid). All lines slope upwards.

Figure 2. Moderating effect of LGO on joy and EEI.

Figure 3
Line graph titled “LGO x JOY” shows the relationship between JOY (x-axis) and EXI (y-axis). Three lines depict LGO at different levels: solid line for +1 SD, dotted line for mean, and dashed line for -1 SD. The lines slope upward, indicating a positive relationship.

Figure 3. Moderating effect of LGO on joy and EXI.

4.3 Robustness checks

To assess the robustness of the main findings, this study re-estimated the structural model by including daily AI usage time as a control variable. As shown in Table 5, the hypothesized relationships remain substantively unchanged. Although the indirect effect of AI usage on EEI becomes marginally significant after controlling for daily AI usage time, the effect size is small and the significance is sensitive to model specification, and is therefore not interpreted as robust. Moreover, daily AI usage time does not exhibit significant direct or indirect effects, suggesting that the main findings are not driven by differences in AI usage intensity.

Table 5
www.frontiersin.org

Table 5. Robustness checks with daily AI usage time as a control variable.

5 Discussion

5.1 Findings

This study examines how AI usage shapes EEI and EXI by integrating the mediating role of joy and the moderating role of LGO. Overall, the findings reveal the differentiated emotional pathways through which AI usage influences distinct forms of innovation, highlighting how the effects of positive emotions depend on task characteristics and individual motivational orientations in AI-enabled work contexts.

The results demonstrate that AI usage significantly promotes both EEI and EXI, suggesting that AI supports employees not only in refining existing processes and resources but also in pursuing new methods and opportunities. These findings align with prior research indicating that AI usage facilitates incremental improvements while simultaneously enabling the generation of novel knowledge through human–AI collaboration (Einola and Khoreva, 2023; Lee and Chung, 2024). Moreover, the results indicate that AI usage contributes to EEI and EXI alongside other individual and contextual factors, consistent with the multi-resource nature of ambidextrous innovation (Boemelburg et al., 2023).

Importantly, the mediating role of joy differs across innovation types. Joy mediates the relationship between AI usage and EXI, but not between AI usage and EEI. This divergence can be understood through the functional alignment between emotional processes and task characteristics. Drawing on the BBT, high-arousal positive emotions such as joy broaden individuals’ momentary thought–action repertoires and facilitate the accumulation of enduring psychological resources over time (Fredrickson, 2001). These functions are especially relevant for EXI, which involves experimentation, novelty seeking, and sustained engagement under conditions of uncertainty and risk (Boemelburg et al., 2023; Mom et al., 2015).

In contrast, EEI involves structured and incremental refinement within established routines, placing relatively weaker demands on emotional arousal (Boemelburg et al., 2023; Mom et al., 2015). In AI-enabled work settings, the functional features of AI, including the recombination of existing process knowledge, can directly support EEI (Scarbrough et al., 2024). Consequently, joy operates as a supplementary rather than a central psychological mechanism in translating AI usage into EEI.

The significant interaction between joy and LGO further clarifies the contingent role of joy in EEI. Joy predicts EEI only among employees with high LGO, indicating that positive affect is more likely to be mobilized into process-focused improvement behaviors when individuals cognitively appraise AI usage as an opportunity for skill development and mastery (Solberg et al., 2022). Under such conditions, joyful experiences associated with AI use may be interpreted as signals of learning progress, thereby sustaining incremental refinement efforts. By contrast, for employees with low LGO, joy is more likely to remain a transient experiential state that does not translate into sustained EEI. This divergence helps explain why the indirect effect of joy in the AI usage–EEI relationship is not significant at the aggregate level.

Taken together, these findings suggest that joy does not function as a universal emotional mechanism linking AI usage to all forms of employee innovation. Instead, its role is contingent on the alignment between emotional functions and task characteristics, as well as on individual motivational orientations. Whereas joy constitutes a core psychological resource for sustaining exploratory innovation, its influence on EEI is conditional and becomes salient only under specific motivational conditions. This pattern is consistent with CAT, which suggests that the implications of emotions for action tendencies vary depending on how individuals appraise situational demands and goal relevance (Lazarus, 1991; So et al., 2015; Tan and Titova, 2024).

5.2 Theoretical and practical implications

5.2.1 Theoretical implications

By addressing two core research objectives, this study advances CAT and the BBT in explaining employee innovation in AI-enabled work contexts.

Objective 1 was to examine whether and how AI usage influences EEI and EXI through discrete positive emotions. The findings show that joy mediates the relationship between AI usage and EXI, but not between AI usage and EEI. This result extends the application of CAT in innovation research by demonstrating that, in AI-enabled contexts, the innovation-enhancing role of joy is contingent on task demands, such that the same emotional experience serves different functions in EEI and EXI. At the same time, this finding refines the BBT by indicating that the broadening potential of positive emotions is not uniformly translated into all forms of innovation. By conceptualizing joy as a discrete positive emotion, the results advance the discrete emotions literature by showing that even high-arousal positive emotions may exert differentiated effects across innovation outcomes.

Objective 2 was to examine when joy translates into EEI and EXI by considering individual motivational orientations. The findings indicate that LGO constitutes a critical boundary condition for the innovation-enhancing role of joy, such that positive emotional experiences are converted into innovative outcomes only under high levels of LGO. This result advances understanding of emotion-innovation linkages by highlighting that the functional consequences of joy depend not only on task demands but also on individuals’ motivational orientations.

Collectively, these findings underscore the interactive roles of cognitive appraisal, discrete emotions, and individual traits in shaping how AI usage translates into employee innovation. By empirically integrating CAT and BBT in an AI-related context, this study provides a more fine-grained theoretical framework for understanding ambidextrous innovation and lays a foundation for future research to explore the differentiated roles of other discrete positive emotions across diverse forms of innovation.

5.2.2 Practical implications

This study offers several theory-informed managerial implications for guiding AI adoption and innovation management.

First, the findings suggest that organizations may benefit from moving beyond a narrow focus on efficiency gains and tool value by designing AI systems and training programs that support learning-oriented experiences rather than solely emphasizing task automation. Specifically, managers may emphasize features such as transparent AI feedback, opportunities for experimentation, and user control over AI-assisted outputs. Prior research suggests that such design choices can help position AI usage as a source of perceived competence development and engagement, which may facilitate the conversion of positive emotional experiences into innovative outcomes.

Second, drawing on prior literature on automation reliance and human-AI interaction, organizations may need to remain attentive to the potential risk of employees’ overreliance on AI in routine and procedural tasks, which has been suggested to undermine active engagement in exploitative innovation. Prior research indicates that reduced initiative in process improvement, weakened task ownership, or uncritical acceptance of AI-generated outputs may signal overreliance. Managers may help ensure that AI functions as an enabling tool rather than a substitute for employees’ problem-solving efforts by adopting practices such as human-in-the-loop designs, reflective review sessions, and accountability mechanisms.

Third, the findings underscore the relevance of LGO, suggesting that AI training and development practices may benefit from differentiation according to employees’ motivational profiles. For employees with high LGO, training programs may place greater emphasis on advanced AI functionalities, exploratory use cases, and open-ended problem-solving tasks that reinforce joy as a signal of learning progress. For employees with low LGO, more structured training approaches, guided learning paths, and targeted feedback may help translate positive affect into sustained innovative behavior. Overall, aligning AI implementation with individual motivational orientations may increase the likelihood of fostering synergy between technological empowerment and employee innovation.

Taken together, these implications should be viewed as plausible and theory-consistent extensions of the findings rather than direct causal conclusions, given the study’s cross-sectional design and modest effect sizes.

5.3 Limitations and directions for future research

This study has several limitations that provide avenues for future research. First, its cross-sectional design restricts the ability to draw strong causal inferences. Future studies could adopt longitudinal designs to better capture the dynamic evolution of AI usage and ambidextrous innovation over time, and conduct experimental studies that manipulate AI-related features (e.g., feedback transparency or autonomy) to strengthen causal inference.

Second, the analysis focused solely on joy as a discrete positive emotion, which may limit the comprehensiveness of the emotional mechanisms examined. Moreover, although joy is conceptually distinct from more stable evaluative states such as job satisfaction in terms of arousal and temporal dynamics, the use of AI as a work tool may blur this distinction in practice. Future studies could therefore explicitly distinguish discrete emotions from general attitudinal constructs, or examine their joint effects, to further clarify the unique role of joy in AI-enabled innovation processes.

Finally, the data were collected primarily from Chinese enterprises, which may constrain the cross-cultural generalizability of the findings. Replications across cultural contexts and multilevel designs examining how organizational AI strategies interact with individual emotions and motivational orientations would further strengthen and extend the present findings.

6 Conclusion

This study investigates the role of discrete emotional and motivational mechanisms in linking AI usage to EEI and EXI. By integrating CAT and the BBT, the findings indicate that joy functions as a differentiated emotional pathway rather than a uniform driver across innovation types. In addition, the study identifies LGO as a critical boundary condition that shapes when joy translates into innovation. Taken together, the findings underscore that the innovation-related consequences of AI usage depend not only on technological features, but also on how employees cognitively appraise AI-enabled work and how emotional experiences interact with individual motivational orientations, offering a key takeaway for research on digital work and AI-enabled innovation.

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.

Ethics statement

The requirement of ethical approval was waived by Institutional Review Board, Dhurakij Pundit University, Thailand for the studies involving humans because Institutional Review Board, Dhurakij Pundit University, Thailand. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

Author contributions

LL: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Writing – original draft, Writing – review & editing. YL: Conceptualization, Formal analysis, Methodology, Writing – review & editing.

Funding

The author(s) declared that financial support was not received for this work and/or its publication.

Conflict of interest

The author(s) declared that this work 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) declared that Generative AI was used in the creation of this manuscript. Solely to improve language clarity and grammar. All ideas, analyses, and conclusions are the responsibility of the authors.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

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.

Supplementary material

The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fhumd.2026.1695355/full#supplementary-material

References

Arnett, J. J. (2023). Joy: an integrative theory. J. Posit. Psychol. 18, 1–14. doi: 10.1080/17439760.2022.2053878

Crossref Full Text | Google Scholar

Bernerth, J. B., and Aguinis, H. (2016). A critical review and best-practice recommendations for control variable usage. Pers. Psychol. 69, 229–283. doi: 10.1111/peps.12103

Crossref Full Text | Google Scholar

Boemelburg, R., Berger, S., Jansen, J. J. P., and Bruch, H. (2023). Regulatory focus climate, organizational structure, and employee ambidexterity: an interactive multilevel model. Hum. Resour. Manag. 62, 701–719. doi: 10.1002/hrm.22155

Crossref Full Text | Google Scholar

Burton, J. W., Lopez-Lopez, E., Hechtlinger, S., Rahwan, Z., Aeschbach, S., Bakker, M. A., et al. (2024). How large language models can reshape collective intelligence. Nat. Hum. Behav. 8, 1643–1655. doi: 10.1038/s41562-024-01959-9,

PubMed Abstract | Crossref Full Text | Google Scholar

Clark, L. A., and Watson, D. (1994). The PANAS-X: Manual for the Positive and Negative Affect Schedule - Expanded Form. Washington, DC: American Psychological Association.

Google Scholar

Da Motta Veiga, S. P., and Turban, D. B. (2014). Are affect and perceived stress detrimental or beneficial to job seekers? The role of learning goal orientation in job search self-regulation. Organ. Behav. Hum. Decis. Process. 125, 193–203. doi: 10.1016/j.obhdp.2014.09.007

Crossref Full Text | Google Scholar

Davenport, T., Guha, A., Grewal, D., and Bressgott, T. (2020). How artificial intelligence will change the future of marketing. J. Acad. Mark. Sci. 48, 24–42. doi: 10.1007/s11747-019-00696-0

Crossref Full Text | Google Scholar

Dennis, A. R., Lakhiwal, A., and Sachdeva, A. (2023). AI agents as team members: effects on satisfaction, conflict, trustworthiness, and willingness to work with. J. Manag. Inf. Syst. 40, 307–337. doi: 10.1080/07421222.2023.2196773

Crossref Full Text | Google Scholar

Denrell, J., Zeijen, A., Romagnoli, M., and Marengo, L. (2025). Absolute, average-based, and rank-based aspirations. Strateg. Manag. J. 46, 1930–1946. doi: 10.1002/smj.3715

Crossref Full Text | Google Scholar

Dweck, C. S., and Leggett, E. L. (1988). A social-cognitive approach to motivation and personality. Psychol. Rev. 95, 256–273. doi: 10.1037/0033-295X.95.2.256

Crossref Full Text | Google Scholar

Einola, K., and Khoreva, V. (2023). Best friend or broken tool? Exploring the co-existence of humans and artificial intelligence in the workplace ecosystem. Hum. Resour. Manag. 62, 117–135. doi: 10.1002/hrm.22147

Crossref Full Text | Google Scholar

Eshraghian, F., Hafezieh, N., Farivar, F., and De Cesare, S. (2024). AI in software programming: understanding emotional responses to GitHub copilot. Inf. Technol. People. 38, 1659–1685. doi: 10.1108/ITP-01-2023-0084

Crossref Full Text | Google Scholar

Forgas, J. P., and George, J. M. (2001). Affective influences on judgments and behavior in organizations: an information processing perspective. Organ. Behav. Hum. Decis. Process. 86, 3–34. doi: 10.1006/obhd.2001.2971

Crossref Full Text | Google Scholar

Fredrickson, B. L. (2001). The role of positive emotions in positive psychology: the broaden-and-build theory of positive emotions. Am. Psychol. 56, 218–226.

Google Scholar

Fredrickson, B. L. (2013). Positive emotions broaden and build. Adv. Exp. Soc. Psychol. 47, 1–53. doi: 10.1016/B978-0-12-407236-7.00001-2

Crossref Full Text | Google Scholar

Guha, A., Bressgott, T., Grewal, D., Mahr, D., Wetzels, M., and Schweiger, E. (2023). How artificiality and intelligence affect voice assistant evaluations. J. Acad. Mark. Sci. 51, 843–866. doi: 10.1007/s11747-022-00874-7

Crossref Full Text | Google Scholar

Hair, J. F., Risher, J. J., Sarstedt, M., and Ringle, C. M. (2019). When to use and how to report the results of PLS-SEM. Eur. Bus. Rev. 31, 2–24. doi: 10.1108/EBR-11-2018-0203

Crossref Full Text | Google Scholar

Hardy, J. H., Day, E. A., North, M. N., and Rockwood, J. (2024). Unpacking on-task effort in performance-based learning: information–knowledge gaps guide effort allocation decisions. J. Appl. Psychol. 109, 77–98. doi: 10.1037/apl0001140

Crossref Full Text | Google Scholar

Hitt, M. A., Beamish, P. W., Jackson, S. E., and Mathieu, J. E. (2007). Building theoretical and empirical bridges across levels: multilevel research in management. Acad. Manag. J. 50, 1385–1399. doi: 10.5465/amj.2007.28166219

Crossref Full Text | Google Scholar

Hwang, Y., and Wu, Y. (2025). The influence of generative artificial intelligence on creative cognition of design students: a chain mediation model of self-efficacy and anxiety. Front. Psychol. 15:1455015. doi: 10.3389/fpsyg.2024.1455015,

PubMed Abstract | Crossref Full Text | Google Scholar

Jatobá, M. N., Ferreira, J. J., Fernandes, P. O., and Teixeira, J. P. (2023). Intelligent human resources for the adoption of artificial intelligence: a systematic literature review. J. Organ. Chang. Manag. 36, 1099–1124. doi: 10.1108/JOCM-03-2022-0075

Crossref Full Text | Google Scholar

Jia, N., Luo, X., Fang, Z., and Liao, C. (2023). When and how artificial intelligence augments employee creativity. Acad. Manag. J. 67, 5–32. doi: 10.5465/amj.2022.0426

Crossref Full Text | Google Scholar

Johnson, M. K. (2020). Joy: a review of the literature and suggestions for future directions. J. Posit. Psychol. 15, 5–24. doi: 10.1080/17439760.2019.1685581

Crossref Full Text | Google Scholar

Kanbach, D. K., Heiduk, L., Blueher, G., Schreiter, M., and Lahmann, A. (2023). The GenAI is out of the bottle: generative artificial intelligence from a business model innovation perspective. Rev. Manag. Sci. 18, 1189–1220. doi: 10.1007/s11846-023-00696-z

Crossref Full Text | Google Scholar

Keltner, D., Sauter, D., Tracy, J., and Cowen, A. (2019). Emotional expression: advances in basic emotion theory. J. Nonverbal Behav. 43, 133–160. doi: 10.1007/s10919-019-00293-3,

PubMed Abstract | Crossref Full Text | Google Scholar

King, P. E. (2020). Joy distinguished: teleological perspectives on joy as a virtue. J. Posit. Psychol. 15, 33–39. doi: 10.1080/17439760.2019.1685578

Crossref Full Text | Google Scholar

Lauenstein, F., Newark, D. A., and Baumann, O. (2025). How mixed performance feedback shapes exploration: the moderating role of self-enhancement. Organ. Sci. 36, 166–185. doi: 10.1287/orsc.2021.15676

Crossref Full Text | Google Scholar

Lazarus, R. S. (1991). Progress on a cognitive-motivational-relational theory of emotion. Am. Psychol. 46, 819–834. doi: 10.1037/0003-066X.46.8.819,

PubMed Abstract | Crossref Full Text | Google Scholar

Lee, B. C., and Chung, J. (2024). An empirical investigation of the impact of ChatGPT on creativity. Nat. Hum. Behav. 8, 1906–1914. doi: 10.1038/s41562-024-01953-1,

PubMed Abstract | Crossref Full Text | Google Scholar

Lench, H. C., Tibbett, T. P., and Bench, S. W. (2016). Exploring the toolkit of emotion: what do sadness and anger do for us? Soc. Personal. Psychol. Compass 10, 11–25. doi: 10.1111/spc3.12229

Crossref Full Text | Google Scholar

Lin, H., and Chen, Q. (2024). Artificial intelligence (AI)-integrated educational applications and college students’ creativity and academic emotions: students and teachers’ perceptions and attitudes. BMC Psychol. 12:487. doi: 10.1186/s40359-024-01979-0,

PubMed Abstract | Crossref Full Text | Google Scholar

Liu, M., Kan, X., Zhao, M., and Li, Q. (2024). Linking perceived anxiety to knowledge purchase likelihood: the mediating role of coping strategies. Curr. Psychol. 43, 13630–13641. doi: 10.1007/s12144-023-05392-5

Crossref Full Text | Google Scholar

Luo, X., Qin, M. S., Fang, Z., and Qu, Z. (2021). Artificial intelligence coaches for sales agents: caveats and solutions. J. Mark. 85, 14–32. doi: 10.1177/0022242920956676

Crossref Full Text | Google Scholar

Malik, A., Budhwar, P., and Mohan, H. (2023). Employee experience-the missing link for engaging employees: insights from an MNE’s AI - based HR ecosystem. Hum. Resour. Manag. 62, 97–115. doi: 10.1002/hrm.22133

Crossref Full Text | Google Scholar

Malik, A., De Silva, M. T. T., Budhwar, P., and Srikanth, N. R. (2021). Elevating talents’ experience through innovative artificial intelligence-mediated knowledge sharing: evidence from an IT-multinational enterprise. J. Int. Manag. 27:100871. doi: 10.1016/j.intman.2021.100871

Crossref Full Text | Google Scholar

Man Tang, P., Koopman, J., McClean, S. T., Zhang, J. H., Li, C. H., De Cremer, D., et al. (2022). When conscientious employees meet intelligent machines: an integrative approach inspired by complementarity theory and role theory. Acad. Manag. J. 65, 1019–1054. doi: 10.5465/amj.2020.1516

Crossref Full Text | Google Scholar

Marshall, D. R., Davis, W. D., Dibrell, C., and Ammeter, A. P. (2019). Learning off the job: examining part-time entrepreneurs as innovative employees. J. Manag. 45, 3091–3113. doi: 10.1177/0149206318779127

Crossref Full Text | Google Scholar

Meade, A. W., and Craig, S. B. (2012). Identifying careless responses in survey data. Psychol. Methods 17, 437–455. doi: 10.1037/a0028085

Crossref Full Text | Google Scholar

Mithas, S., Chen, Z., Saldanha, T. J. V., and De Oliveira Silveira, A. (2022). How will artificial intelligence and industry 4.0 emerging technologies transform operations management? Prod. Oper. Manag. 31, 4475–4487. doi: 10.1111/poms.13864

Crossref Full Text | Google Scholar

Mom, T. J. M., Chang, Y.-Y., Cholakova, M., and Jansen, J. J. P. (2019). A multilevel integrated framework of firm HR practices, individual ambidexterity, and organizational ambidexterity. J. Manag. 45, 3009–3034. doi: 10.1177/0149206318776775

Crossref Full Text | Google Scholar

Mom, T. J. M., Fourné, S. P. L., and Jansen, J. J. P. (2015). Managers’ work experience, ambidexterity, and performance: the contingency role of the work context. Hum. Resour. Manag. 54, s133–s153. doi: 10.1002/hrm.21663

Crossref Full Text | Google Scholar

Noy, S., and Zhang, W. (2023). Experimental evidence on the productivity effects of generative artificial intelligence. Science 381, 187–192. doi: 10.1126/science.adh2586,

PubMed Abstract | Crossref Full Text | Google Scholar

Ocal, A., and Crowston, K. (2024). Framing and feelings on social media: the futures of work and intelligent machines. Inf. Technol. People. 37, 2462–2488. doi: 10.1108/ITP-01-2023-0049

Crossref Full Text | Google Scholar

Otto, B. D., Schuessler, E. S., Sydow, J., and Vogelgsang, L. (2024). Finding creativity in predictability: seizing Kairos in Chronos through temporal work in complex innovation processes. Organ. Sci. 35, 1795–1822. doi: 10.1287/orsc.2020.14743

Crossref Full Text | Google Scholar

Peres, R., Schreier, M., Schweidel, D., and Sorescu, A. (2023). On ChatGPT and beyond: how generative artificial intelligence may affect research, teaching, and practice. Int. J. Res. Mark. 40, 269–275. doi: 10.1016/j.ijresmar.2023.03.001

Crossref Full Text | Google Scholar

Podsakoff, P. M., MacKenzie, S. B., Lee, J.-Y., and Podsakoff, N. P. (2003). Common method biases in behavioral research: a critical review of the literature and recommended remedies. J. Appl. Psychol. 88, 879–903. doi: 10.1037/0021-9010.88.5.879,

PubMed Abstract | Crossref Full Text | Google Scholar

Scarbrough, H., Chen, Y., and Patriotta, G. (2024). The AI of the beholder: intra-professional Sensemaking of an epistemic technology. J. Manag. Stud. 62, 1885–1913. doi: 10.1111/joms.13065

Crossref Full Text | Google Scholar

Schwarz, N. (2012). Feelings-as-information theory. P. LangeVan, A. Kruglanski, and E. Higgins, Handbook of theories of social psychology : Volume 1 (pp. 289–308). London: SAGE Publications Ltd.

Google Scholar

Shao, Y., Huang, C., Song, Y., Wang, M., Song, Y. H., and Shao, R. (2024). Using augmentation-based AI tool at work: a daily investigation of learning-based benefit and challenge. J. Manag. 51, 3352–3390. doi: 10.1177/01492063241266503

Crossref Full Text | Google Scholar

Smith, C. A., and Lazarus, R. S. (1993). Appraisal components, core relational themes, and the emotions. Cognit. Emot. 7, 233–269.

Google Scholar

So, J., Achar, C., Han, D., Agrawal, N., Duhachek, A., and Maheswaran, D. (2015). The psychology of appraisal: specific emotions and decision-making. J. Consum. Psychol. 25, 359–371. doi: 10.1016/j.jcps.2015.04.003

Crossref Full Text | Google Scholar

Solberg, E., Sverdrup, T. E., Sandvik, A. M., and Schei, V. (2022). Encouraging or expecting flexibility? How small business leaders’ mastery goal orientation influences employee flexibility through different work climate perceptions. Hum. Relat. 75, 2246–2271. doi: 10.1177/00187267211042538

Crossref Full Text | Google Scholar

Spring, M., Faulconbridge, J., and Sarwar, A. (2022). How information technology automates and augments processes: insights from artificial-intelligence-based systems in professional service operations. J. Oper. Manag. 68, 592–618. doi: 10.1002/joom.1215

Crossref Full Text | Google Scholar

Sun, J., Wayne, S. J., and Liu, Y. (2022). The roller coaster of leader affect: an investigation of observed leader affect variability and engagement. J. Manag. 48, 1188–1213. doi: 10.1177/01492063211008974

Crossref Full Text | Google Scholar

Tan, C. S., and Titova, L. (2024). Enjoying the moment of joy: culture and self during emotional experience. J. Posit. Psychol., 1–12. doi: 10.1080/17439760.2024.2387338

Crossref Full Text | Google Scholar

To, M. L., Fisher, C. D., and Ashkanasy, N. M. (2015). Unleashing angst: negative mood, learning goal orientation, psychological empowerment and creative behaviour. Hum. Relat. 68, 1601–1622. doi: 10.1177/0018726714562235

Crossref Full Text | Google Scholar

Vandewalle, D. (1997). Development and validation of a work domain goal orientation instrument. Educ. Psychol. Meas. 57, 995–1015.

Google Scholar

Verma, S., and Singh, V. (2022). Impact of artificial intelligence-enabled job characteristics and perceived substitution crisis on innovative work behavior of employees from high-tech firms. Comput. Hum. Behav. 131:107215. doi: 10.1016/j.chb.2022.107215

Crossref Full Text | Google Scholar

Welpe, I. M., Spörrle, M., Grichnik, D., Michl, T., and Audretsch, D. B. (2012). Emotions and opportunities: the interplay of opportunity evaluation, fear, joy, and anger as antecedent of entrepreneurial exploitation. Entrep. Theory Pract. 36, 69–96. doi: 10.1111/j.1540-6520.2011.00481.x

Crossref Full Text | Google Scholar

Williams, L. J., Hartman, N., and Cavazotte, F. (2010). Method variance and marker variables: a review and comprehensive CFA marker technique. Organ. Res. Methods 13, 477–514. doi: 10.1177/1094428110366036

Crossref Full Text | Google Scholar

Zhang, Y., DeCarlo, T. E., Manikas, A. S., and Bhattacharya, A. (2023). To exploit or explore? The impact of crowdfunding project descriptions and backers’ power states on funding decisions. J. Acad. Mark. Sci. 51, 444–462. doi: 10.1007/s11747-022-00871-w

Crossref Full Text | Google Scholar

Zhou, Y., Wang, L., and Chen, W. (2023). The dark side of AI-enabled HRM on employees based on AI algorithmic features. J. Organ. Chang. Manag. 36, 1222–1241. doi: 10.1108/JOCM-10-2022-0308

Crossref Full Text | Google Scholar

Keywords: ambidexterity, artificial intelligence usage, employees’ exploitative innovation, employees’ exploratory innovation, joy, learning goal orientation, PLS-SEM, positive emotions

Citation: Liu L and Lv Y (2026) How AI usage shapes employees’ exploitative and exploratory innovation: joy as a mediator and learning goal orientation as a moderator. Front. Hum. Dyn. 8:1695355. doi: 10.3389/fhumd.2026.1695355

Received: 29 August 2025; Revised: 18 January 2026; Accepted: 22 January 2026;
Published: 05 February 2026.

Edited by:

Marco De Angelis, University of Bologna, Italy

Reviewed by:

Carlos Santiago-Torner, University of Vic - Central University of Catalonia, Spain
Fan Chuanhao, Hohai University, China

Copyright © 2026 Liu and Lv. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

*Correspondence: Li Liu, am95bHkwODA3QDE2My5jb20=

Disclaimer: 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.