Human-AI collaboration often underperforms due to a lack of motivation-supportive system design. This paper proposes a framework grounded in work design theory – specifically the Job Characteristics Model (JCM) – to guide the development and evaluation of AI systems. We introduce qualitative evaluation anchors that translate core job design criteria into assessable aspects of AI-supported work. These anchors were developed through a theory-driven process that combines work design theory with recent literature on AI’s impact on work characteristics. The goal is to foster intrinsically motivating and cognitively engaging human roles in AI collaboration, thereby enhancing overall human-AI team performance.
While Artificial Intelligence (AI) technologies continue to evolve, human-AI collaboration in professional contexts often underperforms (Vaccaro et al., 2024). In many cases, humans tend to engage only superficially with AI-generated outputs, resulting in suboptimal decision quality (Buçinca et al., 2024). A key issue is that current AI systems are designed under the assumption that humans are naturally motivated to engage with AI outputs (Buçinca, 2024). However, motivation is not a fixed trait, but shaped by work design (Hackman & Oldham, 1976; Morgeson & Humphrey, 2006; Parker & Grote, 2022). Many AI systems reduce human roles to merely accepting or rejecting outputs, which fosters passive human behavior, impedes learning and skill development, and can lead to overreliance on automation (Buçinca et al., 2021, 2025; Endsley, 2023; Parker & Grote, 2022).
To avoid this, the design of AI systems must actively support psychologically founded prerequisites that promote intrinsic motivation. This paper explores how such job design principles can be systematically integrated into AI design and evaluation. We propose theoretically grounded job design criteria and outline their qualitative operationalization for human- AI collaboration, as currently conceptualized within the HORIZON project AI4REALNET.
Chapter 2 outlines key challenges in current AI design. Chapter 3 presents the theoretical framework based on the Job Characteristics Model. Chapter 4 details the evaluation anchors as conceptual results. Chapter 5 discusses limitations and future directions.
Challenges in Current AI Design Despite advances in AI, joint human-AI decision-making often underperforms compared to the best-performing individual, whether human or AI (Vaccaro et al., 2024). Endsley’s (2023) Ironies of AI highlight how features intended to enhance system performance paradoxically undermine it. Many AI systems are opaque, making it difficult for humans to understand how outputs are generated. Simultaneously, humans tend to overestimate AI capabilities, leading to overreliance, where humans trust in AI outputs, even when suboptimal (Endsley, 2023). This amplifies automation bias – the tendency of humans to uncritically accept automation-generated outputs (Buçinca et al., 2021; Lee & See, 2004; Parasuraman & Manzey, 2010). Such effects are particularly problematic in professional high-stake contexts. Over time, they can impair learning and skill development, ultimately leading to deskilling (Buçinca, 2024; Buçinca et al., 2025).
Providing explainable AI (XAI) and interpretable models is a common response to these challenges (Schaffer et al., 2019), but remains insufficient as humans tend not to engage with it (Buçinca et al., 2024). One main reason for this lack of human engagement is the inappropriate allocation of tasks between humans and AI (Joe et al., 2015; Pritchett et al., 2014). Many systems allocate tasks based on technical feasibility rather than human strengths (Joe et al., 2015; Pritchett et al., 2014), leading to so-called ≪left-over≫roles (Bailey, 1989; Parker & Grote, 2022), like system monitoring or handling exceptions (Feigh & Pritchett, 2014). These roles are cognitively demanding and psychologically unsuitable (Feigh & Pritchett, 2014), particularly as humans are ill-equipped to continuously monitor opaque, complex systems without active involvement (Yampolskiy, 2025).
To address this, AI design must move beyond a purely technology- driven approach and instead integrate psychologically founded principles that support intrinsic motivation to engage with AI. Therefore, we propose a work-design-based approach that ensures AI systems foster meaningful and cognitively engaging roles for humans. This approach builds on established job design theories and is outlined in detail in the following chapter.
THEORETICAL FRAMEWORK – THE JCM AND ITS EXTENSION The Job Characteristics Model (JCM) by Hackman and Oldham (1976) defines five core task characteristics – skill variety, task identity, task significance, autonomy, and feedback – that foster critical psychological states such as experienced meaningfulness, experienced responsibility, and knowledge of results. These criteria are particularly relevant in human-AI collaboration, as AI systems shape how tasks are structured and executed (Parker & Grote, 2022; Zhang et al., 2025), ultimately influencing motivation (Buçinca et al., 2024; Liu & Li, 2025).
To address the cognitive demands of AI-supported, knowledge-intensive work, we draw on Morgeson & Humphrey’s (2006) extension of the
Applying Job Design Criteria for Effective Human-AI Collaboration 13
JCM, which includes knowledge characteristics such as job complexity, information processing, problem-solving, and specialization. Notably, they reclassified skill variety as a knowledge characteristic and complemented the task characteristics with task variety. These criteria reflect the degree of humans’ active cognitive engagement – a key factor in preserving expertise and preventing deskilling over time (Buçinca et al., 2025; Liu & Li, 2025; Parker & Grote, 2022).
To translate these criteria into a practical tool for the design and evaluation of AI, we operationalized them as qualitative anchors in a theory-driven process: starting from core job design literature, we conducted a targeted review on how AI affects each characteristic and synthesized the insights. At this stage, however, the framework is conceptual and requires empirical validation. Semi-structured interview guides and observational protocols to validate these anchors will be developed at a later stage.
The following section presents the task characteristics, their relevance in the human-AI context and the corresponding anchors, followed by knowledge characteristics in the same logic.
EVALUATION ANCHORS FOR HUMAN-AI COLLABORATION Task Characteristics
Task Significance is defined as “the degree to which the job has a substantial impact on the lives or work of other people, whether in the immediate organization or external environment” (Hackman & Oldham, 1976, p. 257). In AI-supported environments, perceived task significance may be diminished if meaningful aspects of a task are automated (Liu & Li, 2025). However, when AI takes over routine tasks, humans may focus more on value-adding activities. Table 1 presents anchors that capture how humans perceive both the significance of their own and the AI’s contribution.
Table 1: Evaluation anchors for task significance.
Table available in the original PDF paper.
High Medium Low
Humans do not recognize
Humans recognize both,
Humans have a basic
their own impact and the impact of AI on the overall performance, including the impact on others. AI makes this transparent. The task is perceived as significant for the overall performance.
understanding of how their actions impact the overall performance, but the impact is not always obvious. AI makes it partly transparent. The task is perceived as partly significant for the overall performance.
the impact of their own work or the AI on the overall performance. AI is not able to make impacts transparent. The task is perceived as routine and without significant importance for the overall performance.
Task Identity refers to “the extent to which a job requires completing a ‘whole’ and identifiable piece of work”(Hackman & Oldham, 1976, p. 257). In AI-supported settings, workflows are often fragmented as automation takes over individual steps. As a result, humans may only engage with isolated
parts of the process, making it harder to understand how their contribution fits into the final outcome (Liu & Li, 2025). When human roles are limited to monitoring or validating AI outputs, their connection to a complete task may weaken (Feigh & Pritchett, 2014; Joe et al., 2015; Parker & Grote, 2022). Table 2 presents evaluation anchors that assess the extent to which humans perceive their work as coherent and complete.
Table 2: Evaluation anchors for task identity.
Table available in the original PDF paper.
High Medium Low Humans are assigned a Humans are involved in Humans only perform selected task steps. AI handles key steps and makes contribution partially transparent.
isolated actions without seeing how their input fits into a complete task. AI executes most of the process.
complete task (including planning, preparation, execution and follow-up tasks) and see how their work contributes to the whole. AI supports without fragmenting the process and by making contributions transparent.
Task Variety is about performing a range of different activities (Morgeson & Humphrey, 2006; Stegmann et al., 2010). AI can either increase variety by taking over routine tasks or reduce it by narrowing human roles to repetitive functions (Brougham & Haar, 2018; Liu & Li, 2025). Table 3 provides evaluation anchors to assess whether humans engage in diverse activities or if AI restricts task variety.
Table 3: Evaluation anchors for task variety.
Table available in the original PDF paper.
High Medium Low
Humans mainly carry out
Humans are involved in
Humans are confronted
various tasks, but these are only partially interrelated. AI supports task variety but lacks context or integration between tasks.
with different types of problems, decision-making situations and contextual conditions that require different strategies for action.
unrelated tasks. Humans are mainly occupied with monitoring the AI, executing its instructions or selecting AI-generated options in a limited number of tasks.
Autonomy refers to the degree of freedom individuals have in scheduling their work, making decisions, and choosing how to perform tasks (Hackman & Oldham, 1976; Morgeson & Humphrey, 2006; Stegmann et al., 2010). In AI-supported work environments, autonomy may be constrained when systems enforce workflows or prescribe decisions (Feigh & Pritchett, 2014; Levy et al., 2021; Liu & Li, 2025; Parker & Grote, 2022;
Applying Job Design Criteria for Effective Human-AI Collaboration 15
Schaap et al., 2023). At the same time, AI can also enhance autonomy by offering flexible, human-adaptable options. The following anchors in Table 4 reflect the extent to which humans retain autonomy when working with AI systems.
Table 4: Evaluation anchors for autonomy.
Table available in the original PDF paper.
High Medium Low
Planning: The timing and
Planning: People have a
Planning: Humans have
sequence of human tasks are largely or completely determined by AI. Humans have hardly any scope to change these. Decision-making: The AI
complete control over the planning of the timing and sequence of their tasks. The AI is flexible and adaptable to individual plans. Decision-making:
certain amount of leeway when planning the time and sequence of their tasks. The AI’s specifications regarding time and sequence can be partially adapted by humans. Decision-making:
dedicates decisions or enforces predefined options. Humans have little or no say in implementing alternative or novel solutions, even if they have them. Method: The AI enforces
Humans are free to implement their own ideas or AI-supported solutions. They can make independent decisions on goals and courses of action. The AI can be fully overridden. Method: Humans have
Humans choose between a limited number of AI-generated options. Their ability to implement self-developed ideas is restricted but not entirely blocked. Method: Humans can
a fixed way of completing a task. Humans follow predefined steps with no flexibility in choosing or adapting the strategy.
adjust their approach within a set of predefined methods to carry out the task. The AI allows limited customization.
complete freedom to decide which strategy to use to complete the task. The AI is adaptable to different methods and individual preferences.
Feedback refers to “the degree to which carrying out the work activities required by the job results in the individual obtaining direct and clear information about the effectiveness of his or her performance” (Hackman & Oldham, 1976, p. 258). AI can enhance feedback through immediate, data-driven responses but may also present it in ways that are difficult to understand (Shin et al., 2024). Table 5 presents evaluation anchors that assess the clarity, immediacy, and comprehensibility of AI-generated feedback.
Table 5: Evaluation anchors for feedback.
Table available in the original PDF paper.
High Medium Low Humans receive direct, Humans receive direct
The feedback from the AI
comprehensible and prompt feedback from the AI regarding their own performance.
is not comprehensible or timely, which makes it difficult to assess their own performance.
feedback from AI regarding their own performance, but it is only partially comprehensible and/or not prompt.
Knowledge Characteristics Skill Variety refers to the breadth of skills required to perform a job (Hackman & Oldham, 1976; Morgeson & Humphrey, 2006; Stegmann et al., 2010). Depending on whether AI complements or replaces human skills, it can either increase skill variety by requiring the integration of diverse skills such as analytical and interpersonal skills, or reduce it by narrowing tasks to highly specialized routines (Liu & Li, 2025; Parker & Grote, 2022). Table 6 presents evaluation anchors that assess whether the AI-supported work context demands a broad range of human skills or restricts the variety of skills required.
Table 6: Evaluation anchors for skill variety.
Table available in the original PDF paper.
High Medium Low The humans’ task does The humans’ task requires The humans’ task requires a broad range of relevant knowledge and different relevant skills.
not require broad knowledge and different skills. The tasks are very specialized.
a narrower range of skills and knowledge compared to without AI.
Problem Solving refers to “the degree to which a job requires unique ideas or solutions” (Morgeson & Humphrey, 2006, p. 1323). While AI can support decision-making by offering insights and alternatives (Levy et al., 2021; Salimzadeh & Gadiraju, 2024), it may also foster overreliance and reduce opportunities for independent problem-solving (Buçinca et al., 2021; Schaffer et al., 2019; Vaccaro et al., 2024). As a result, humans risk becoming passive recipients rather than active contributors, which may lead to a gradual loss of expertise (Buçinca, 2024; Buçinca et al., 2025; Parker & Grote, 2022). Table 7 presents evaluation anchors that assess the extent to which individuals can independently develop solutions versus relying on predefined AI-generated options.
Table 7: Evaluation criteria for problem solving.
Table available in the original PDF paper.
High Medium Low The humans develop their The development of Humans have little own ideas or solutions and can react creatively to unforeseen situations. AI does not provide solutions.
human ideas or solutions is restricted because the AI proposes solution options. However, these can be adapted.
opportunity to develop or apply their own ideas or solutions. AI provides fixed solutions with minimal room for customization.
Job Complexity refers to the degree to which a task is mentally demanding or difficult (Morgeson & Humphrey, 2006; Stegmann et al., 2010). While AI may simplify operational tasks, it can simultaneously increase cognitive load through requirements such as monitoring, exception handling, and adaption to AI-generated outputs (Buçinca et al., 2025; Feigh & Pritchett, 2014; Levy et al., 2021). In this way, AI tends to shift complexity rather than
Applying Job Design Criteria for Effective Human-AI Collaboration 17
reduce it. Table 8 presents evaluation anchors that assess the cognitive demands placed on humans.
Table 8: Evaluation anchors for job complexity.
Table available in the original PDF paper.
High Medium Low Humans engage in Humans apply some The task is cognitively cognitive effort, mainly through reviewing AI suggestions or applying predefined rules.
under- or overchallenging. Humans either passively monitor AI outputs or face results that are too complex to interpret. In both cases, the task lacks constructive cognitive activation.
sustained mental effort, applying judgment, practice, and adaptive thinking. The complexity stems from the task itself and is experienced constructive challenging. AI provides clear and relevant output without unnecessary mental effort.
Information Processing refers to “the degree to which a job requires attending to and processing data or other information” (Morgeson & Humphrey, 2006, p. 1323). AI can support this function by pre-filtering data, but may also undermine it by fully pre-processing information, turning humans into passive recipients (Buçinca, 2024). As such, the human role in information processing is a critical factor in maintaining cognitive engagement and preserving expertise. Table 9 presents evaluation anchors that assess the extent of human involvement in interpreting and integrating information.
Table 9: Evaluation anchors for information processing.
Table available in the original PDF paper.
High Medium Low Humans process Humans consume fully Humans actively search, pre-filtered information selected by the AI. Some interpretation is needed, but the system steers attention and framing.
processed AI outputs. Information is final, and little to no analysis or interpretation is required.
analyse, and integrate information. AI provides structured but unprocessed data and may support understanding through simulations. Interpretation remains a human task.
Specialization refers to “the extent to which a job involves performing specialized tasks or possessing specialized knowledge and skill” (Morgeson & Humphrey, 2006, p. 1324) and reflects whether human expertise remains essential or is increasingly replaced by automated processes in AI-supported work environments. AI may foster specialization by taking
over routine tasks, enabling humans to concentrate on expert-level activities (Brougham & Haar, 2018; Liu & Li, 2025). However, this criterion is critical when it results in low skill variety. Table 10 presents evaluation anchors that assess whether human tasks require in-depth domain expertise to perform their tasks or whether AI systems reduce the need for specialization.
Table 10: Evaluation anchors for specialization.
Table available in the original PDF paper.
High Medium Low The task requires in-depth A certain amount of
Specialized knowledge is
knowledge in a specific area to perform the task effectively. AI handles routine work.
specialized knowledge and skills are required, but most of the time general knowledge is sufficient.
not required. General knowledge is sufficient, as AI handles the expertise by solving problems.
Together, these evaluation anchors represent an initial step toward a structured conceptual framework for assessing whether AI systems support or hinder motivational and cognitive engagement in human-AI collaboration.
CONCLUSION
Despite technological advancements, human-AI collaboration continues to face critical challenges such as overreliance (Endsley, 2023), automation bias (Buçinca et al., 2021; Lee & See, 2004; Parasuraman & Manzey, 2010), and poorly aligned task allocation that often overlooks human strengths and weaknesses (Joe et al., 2015; Pritchett et al., 2014). Current solutions like XAI fail to address the root problem: the lack of motivation-supportive interaction design, which is a prerequisite for meaningful human engagement with AI systems (Buçinca et al., 2024). To address this, we introduce an initial conceptual step toward a design and evaluation framework grounded in work design theory. Specifically, we apply and adapt criteria from the Job Characteristics Model (Hackman & Oldham, 1976) and its extension (Morgeson & Humphrey, 2006) to the context of AI-supported work. The resulting evaluation anchors offer a structured approach to assessing key motivational job characteristics in AI system design.
LIMITATIONS AND OUTLOOK A key limitation of this work is the lack of empirical validation of the evaluation anchors. To address this, semi-structured interview guides and observational protocols are currently being developed to operationalize and test the framework in real-world settings.
ACKNOWLEDGMENT
AI4REALNET has received funding from European Union’s Horizon Europe Research and Innovation programme under the Grant Agreement No 101119527 and from the Swiss State Secretariat for Education, Research and lnnovation (SERI). Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union
Applying Job Design Criteria for Effective Human-AI Collaboration 19
and SERI. Neither the European Union nor the granting authority can be held responsible for them.
REFERENCES
Bailey, R. W. (1989). Human performance engineering (2. Aufl.). Prentice-Hall. Brougham, D., & Haar, J. (2018). Smart Technology, Artificial Intelligence, Robotics, and Algorithms (STARA): Employees’ perceptions of our future workplace. Journal of Management & Organization, 24(2), 239–257. https:// doi.org/10.1017/jmo.2016.55 Buçinca, Z. (2024). Optimizing Decision-Maker’s Intrinsic Motivation for Effective
Human-AI Decision-Making. Extended Abstracts of the CHI Conference on Human Factors in Computing Systems, 1–5. https://doi.org/10.1145/ 3613905.3638179 Buçinca, Z., Malaya, M. B., & Gajos, K. Z. (2021). To Trust or to Think: Cognitive Forcing Functions Can Reduce Overreliance on AI in AI-assisted Decision-making. Proceedings of the ACM on Human-Computer Interaction, 5(CSCW1), 1–21. https://doi.org/10.1145/3449287 Buçinca, Z., Swaroop, S., Paluch, A. E., Doshi-Velez, F., & Gajos, K. Z. (2025).
Contrastive Explanations That Anticipate Human Misconceptions Can Improve Human Decision-Making Skills (No. arXiv:2410.04253). arXiv. https://doi.org/ 10.48550/arXiv.2410.04253 Buçinca, Z., Swaroop, S., Paluch, A. E., Murphy, S. A., & Gajos, K. Z. (2024).
Towards Optimizing Human-Centric Objectives in AI-Assisted Decision-Making With Offline Reinforcement Learning (No. arXiv:2403.05911). arXiv. https:// doi.org/10.48550/arXiv.2403.05911 Endsley, M. R. (2023). Ironies of artificial intelligence. Ergonomics, 66(11), 1656–1668. https://doi.org/10.1080/00140139.2023.2243404 Feigh, K. M., & Pritchett, A. R. (2014). Requirements for Effective Function
Allocation: A Critical Review. Journal of Cognitive Engineering and Decision Making, 8(1), 23–32. https://doi.org/10.1177/1555343413490945 Hackman, J. R., & Oldham, G. R. (1976). Motivation through the design of work: Test of a theory. Organizational Behavior and Human Performance, 16(2), 250–279. https://doi.org/10.1016/0030-5073(76)90016-7 Joe, J. C., O’Hara, J., Hugo, J. V., & Oxstrand, J. H. (2015). Function Allocation
for Humans and Automation in the Context of Team Dynamics. Procedia Manufacturing, 3, 1225–1232. https://doi.org/10.1016/j.promfg.2015.07.204 Lee, J. D., & See, K. A. (2004). Trust in Automation: Designing for Appropriate Reliance. Human Factors. Levy, A., Agrawal, M., Satyanarayan, A., & Sontag, D. (2021). Assessing
the Impact of Automated Suggestions on Decision Making: Domain Experts Mediate Model Errors but Take Less Initiative. Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems, 1–13. https://doi.org/ 10.1145/3411764.3445522 Liu, X., & Li, Y. (2025). Examining the Double-Edged Sword Effect of AI Usage on Work Engagement: The Moderating Role of Core Task Characteristics Substitution. Behavioral Sciences, 15(2), 206. https://doi.org/ 10.3390/bs15020206 Morgeson, F. P., & Humphrey, S. E. (2006). The Work Design Questionnaire (WDQ): Developing and validating a comprehensive measure for assessing job design and the nature of work. Journal of Applied Psychology, 91(6), 1321–1339. https:// doi.org/10.1037/0021-9010.91.6.1321
Parasuraman, R., & Manzey, D. H. (2010). Complacency and Bias in Human Use of Automation: An Attentional Integration. Human Factors: The Journal of the Human Factors and Ergonomics Society, 52(3), 381–410. https://doi.org/ 10.1177/0018720810376055 Parker, S. K., & Grote, G. (2022). Automation, Algorithms, and Beyond: Why Work
Design Matters More Than Ever in a Digital World. Applied Psychology, 71(4), 1171–1204. https://doi.org/10.1111/apps.12241 Pritchett, A. R., Kim, S. Y., & Feigh, K. M. (2014). Measuring Human-Automation Function Allocation. Journal of Cognitive Engineering and Decision Making, 8(1), 52–77. https://doi.org/10.1177/1555343413490166 Salimzadeh, S., & Gadiraju, U. (2024). When in Doubt! Understanding the Role of
Task Characteristics on Peer Decision-Making with AI Assistance. Proceedings of the 32nd ACM Conference on User Modeling, Adaptation and Personalization, 89–101. https://doi.org/10.1145/3627043.3659567 Schaap, G., Bosse, T., & Hendriks Vettehen, P. (2023). The ABC of algorithmic aversion: Not agent, but benefits and control determine the acceptance of automated decision-making. AI & SOCIETY. https://doi.org/10.1007/s00146- 023-01649-6 Schaffer, J., O’Donovan, J., Michaelis, J., Raglin, A., & Höllerer, T. (2019). I can
do better than your AI: Expertise and explanations. Proceedings of the 24th International Conference on Intelligent User Interfaces, 240–251. https://doi.org/ 10.1145/3301275.3302308 Shin, H., Park, J., Yu, J., Kim, J., Kim, H. Y., & Oh, C. (2024). Looping In: Exploring Feedback Strategies to Motivate Human Engagement in Interactive Machine Learning. International Journal of Human–Computer Interaction, 1–18. https://doi.org/10.1080/10447318.2024.2413293 Stegmann, S., Dick, R. V., Ullrich, J., Charalambous, J., Menzel, B., Egold, N.,
& Wu, T. T.-C. (2010). Der Work Design Questionnaire: Vorstellung und erste Validierung einer deutschen Version. Zeitschrift für Arbeits- und Organisation- spsychologie A&O, 54(1), 1–28. https://doi.org/10.1026/0932-4089/a000002 Vaccaro, M., Almaatouq, A., & Malone, T. (2024). When combinations of humans and AI are useful: A systematic review and meta-analysis. Nature Human Behaviour, 8(12), 2293–2303. https://doi.org/10.1038/s41562-024-02024-1 Yampolskiy, R. V. (2025). On monitorability of AI. AI and Ethics, 5(1), 689–707.
https://doi.org/10.1007/s43681-024-00420-x Zhang, Q., Liao, G., Ran, X., & Wang, F. (2025). The Impact of AI Usage on Innovation Behavior at Work: The Moderating Role of Openness and Job Complexity. Behavioral Sciences, 15(4), 491. https://doi.org/10.3390/bs15040491
LISTED IN THE AHFE DIGITAL BOOKSTORE CATALOGUE
Explore the complete digital volume
Human Interaction and Emerging Technologies (IHIET 2025)
Read this individual research paper free through the original publisher, or explore the full digital proceedings book.
ISBN 978-1-964867-73-1
Explore this ISBN in the AHFE Bookstore ↗Listed in the July 2026 bookstore catalogue; search by ISBN to confirm current availability · Digital book purchase optional