Blessed be intelligent assistance systems at high task rotation? The effect on motivational work design in assembly

We aimed to provide causal evidence on the contradictory effects of projection‐based intelligent assistance systems (IASs) for nine motivational work characteristics (MWCs). IASs are increasingly implemented in assembly to counteract rising cognitive workload due to individualized manufacturing processes. However, how IASs enhance or restrict MWCs is largely unknown. We conducted two studies with experimental vignette methodology. In Study 1 (N1 = 169 German employees), we manipulated an assembly workplace (with IAS vs. without IAS) and tested whether findings indicating only positive effects of IASs in the support of a simple assembly process can be transferred to more complex assembly processes. In Study 2 (N2 = 176 German employees), we manipulated again the assembly workplace (with IAS vs. without IAS) and in addition the dynamic of product changes (task rotation after 1 h vs. no task rotation). Analyzing the data with SPSS 27, we found increased feedback from job and information processing and decreased work scheduling, decision‐making, and work methods autonomy when working with IAS. In Study 2, we did not find the main or interaction effects of task rotation on MWCs. Our experimental evidence suggests that working with IASs represents a double‐edged sword regarding MWCs and that the effect of task rotation is limited. Hence, our results provide vital theoretical implications for a much‐needed work design theory that delineates how new technologies shape work design and practical implications for modern assembly.

example, on the one hand, Walczok and Bipp (2023) recently found that IASs increase feedback from job and information processing, and thus could have a motivation-enhancing effect.On the other hand, findings of qualitative studies (e.g., Blumberg & Kauffeld, 2020) suggest the risk of restricted autonomy and systematic deskilling when working with IASs.
Therefore, the aim of our current study is twofold.First, we aim to provide much-needed empirical evidence for the positive and/or negative effects of IASs on MWCs in complex assembly processes (Study 1).Second, we investigate the interaction effect of IASs and task rotation on MWCs as IASs are increasingly important in highly dynamic work environments characterized by frequent product changes (Study 2).We contribute to the postulation of work design theories that emphasize the role of new technologies as determinants of work characteristics and their interaction with established work design interventions (Gagné et al., 2022).By highlighting human factors in the implementation of innovative technologies, we point out the positive and negative effects of IAS before they are widely implemented in practice.Hence, we help organizations to react to anticipated changes when implementing IASs in assembly, enabling motivating workplaces in smart factories of the future by amplifying the user-centered design of those systems (Stockinger et al., 2021).

| A model of the future of work design
Even though the effects of technologies on work design are evident, comprehensive theories that describe and explain the impact of technological changes in the workplace on work design are still lacking (Gagné et al., 2022;Walczok & Bipp, 2023).Recently, Gagné et al. (2022) propose a novel model, stating that technological changes in the workplace directly affect the design of work by either in-or decreasing MWCs, leading to psychological need (for autonomy, competence, relatedness) satisfaction, and therefore (indirectly) affecting work motivation.As findings on the impact of technologies on work design are inconsistent, they postulate that "there is no deterministic relationship between technology and work design; instead, the effect of new technology on work design, and hence on motivation, depends on various moderating factors" (p.383).They suggest that technology design and organizational implementation factors moderate the direct effect of technological changes on work design, hence leading to positive or negative motivational effects.On the one hand, technologies have the potential to replace routine cognitive tasks and automate "dull, dangerous, and dirty" (Walsh & Strano, 2018, p. XIX) work.The remaining tasks are characterized by high standardization, including increased monitoring of technologies, and lack of job autonomy, task variety, and skill variety (Blumberg & Kauffeld, 2020;Gagné et al., 2022;Parker & Grote, 2022).On the other hand, technologies can also have a positive impact on MWCs by automating simple tasks, resulting in the execution of more complex tasks, increasing job complexity, skill variety, and problem solving (Gagné et al., 2022;Waschull et al., 2020).
Given that a direct test of the theoretical suggestions of Gagné et al. (2022) is missing so far, we explicate the effect of the previously presented IASs on MWCs in assembly by taking into account technological design factors of IASs as well as frequent product changes as organizational implementation factor into account (Figure 1).In detail, we relied on MWCs (task and knowledge characteristics) that are outlined in the work design questionnaire (WDQ) as a comprehensive model of work characteristics (Morgeson & Humphrey, 2006).We focus on task and knowledge characteristics as meta-analytical findings indicate positive associations of these MWCs with favorable work outcomes (e.g., work motivation) (Humphrey et al., 2007) (see Table 1 for definitions and exemplary items of investigated MWCs).Beyond cross-sectional study evidence on the relationship between MWCs and work outcomes, the results of a systematic review by Knight and Parker (2021) demonstrate that work redesign interventions (like task rotation) lead to changes in employees' perceptions of MWCs and motivation with positive downstream effects on job performance.To prevent unintended negative effects of the implementation of new technologies at work, MWCs need to be considered throughout the whole digitalization process, starting with a user-centered development of these technologies (Mlekus et al., 2022;Stockinger et al., 2021).
The research model based on parts of the theoretical model by Gagné et al. (2022).

| Technology design factors of an IAS in assembly
IASs represent a broad range of technical systems from wearables (Javdan et al., 2023), robotic exoskeletons (Luger et al., 2023), virtual or augmented reality (Marklin et al., 2022) to projection-based assistance systems (Mark et al., 2022).Different classification systems exist for the categorization of IASs, which subdivide those systems either in terms of the level of support or demand, type of support, and objectives (Apt et al., 2018), or attribute categories and capability parameters (Mark et al., 2022).In the current study, we investigate the effects of a cognitive-assistive projection-based IAS with a low level of support which is used for quality assessment and verification of assembly steps.
By presenting context-sensitive in situ projections on the work surface, the projection-based assistance system is intended to guide and cognitively relieve assembly workers during assembly processes.With the help of feedback symbols, including a green tick and a red cross, the IAS gives immediate feedback to the worker when an assembly step is performed correctly or incorrectly, respectively.Machine learning allows the IAS to intelligently and automatically learn new assembly sequences for already known and new, varying products (Walczok & Bipp, 2023;Jung et al., 2022).The technical components of this IAS are discussed in detail by Jung et al. (2022).With this set of functions, the IAS represents a common IAS in industrial applications (Jetter et al., 2018).
The use of language-independent instruction material is supposed to facilitate the inclusion of, for example, nonnative speakers, persons with low skill levels, or with cognitive deficits into the labor market (Apt et al., 2018;Mark et al., 2019).Nevertheless, IASs are also intended to be used in the long term by assembly workers after the learning phases, especially in very dynamic work environments characterized by frequent changes of products (such as modern assembly, in which products rotate resulting from individual customer requests).In this way, those systems should ensure long-term, cognitive relief for assembly workers (Egger-Lampl et al., 2019).However, this means that IASs transform workplaces by altering MWCs.
The first causal evidence on how a cognitive-assistive IAS modifies task and knowledge characteristics stems from Walczok and Bipp (2023).They presented 203 German and British blue-collar workers with a hypothetical assembly workstation and an exemplary, simple work process (assembly of a box) with or without the support of the IAS.Walczok and Bipp (2023) expected an increase in feedback from job as the IAS provides assembly workers with more and immediate feedback for every assembly step in the form of a green tick or a red cross, respectively.Thus, the feedback function as a technological design factor of the IAS leads to higher feedback from job compared with working without IAS.They postulated contradicting effects on information processing.On the one hand, the cognitive support of assembly workers and the takeover of cognitive tasks by instructing T A B L E 1 Definition and exemplary items for investigated motivational work characteristics (MWCs) based on the work design questionnaire (WDQ).

Task characteristics
Feedback from job "degree to which the job provides direct and clear information about the effectiveness of task performance" "The job itself provides feedback on my performance." Work scheduling autonomy "extent to which a job allows freedom independence, and discretion to schedule work" "The job allows me to plan how I do my work." Decision-making autonomy "extent to which a job allows freedom independence, and discretion to […] make decisions" "The job allows me to make a lot of decisions on my own." Work methods autonomy "extent to which a job allows freedom independence, and discretion to […] choose the methods used to perform tasks" "The job allows me to decide on my own how to go about doing my work"

Information processing
"degree to which a job requires attending to and processing data or other information" "The job requires me to analyze a lot of information."

Job complexity
"extent to which the tasks on a job are complex and difficult to perform" "The job comprises relatively uncomplicated tasks."(r) Problem-solving "degree to which a job requires unique ideas or solutions" "The job requires me to be creative."

Specialization
"extent to which a job involves performing specialized tasks or possessing specialized knowledge and skill" "The job requires a depth of knowledge and expertise."

Skill variety
"extent to which a job requires an individual to use a variety of different skills to complete the work" "The job requires a variety of skills." Notes: (r) = reverse scored.Exemplary items in italics.Definitions (pp. 1323Definitions (pp. -1324) ) and exemplary items (pp.1337-1338) stem from Morgeson and Humphrey (2006).
assembly steps could decrease assembly workers' information processing.On the other hand, higher information processing could be increased due to the necessity that employees need to monitor and process additional information (due to feedback and instruction functions of the IAS) when working with IAS compared with working without the support of IAS.Indeed, Walczk and Bipp (2023) identified increased feedback from job (d = 0.85), and information processing (d = 0.33) when working with IAS compared with working without the support of the IAS, but no differences in other MWCs.These findings suggest the impact of IASs on MWCs is purely positive.
H1. Feedback from job is significantly higher in work with IASs than in work without IASs.
H2. Information processing is significantly higher in work with IASs than in work without IASs.
However, such effects seem contrary to prior qualitative studies which propose negative effects on MWCs, such as reduced autonomy or a systematic deskilling of workers (e.g., Blumberg & Kauffeld, 2020).
Also, Walczok and Bipp (2023) criticize that their findings might be restricted by general low ratings in MWCs (floor effects) with a "highly simplified" (p.12) work process in their study, resulting in limited opportunities for the identification of reduced MWCs due to the IAS.
Hence, to rule out this alternative explanation, we tested for such negative effects of an IAS in a more complex assembly process.Since the instruction function of the IAS as a technological design factor strictly standardizes assembly processes by instructing workers through the projection of upcoming assembly steps, the IAS specifies the order of assembly steps and which tools to use.(Apt et al., 2018).Thus, the instruction function of the IAS can lead to the takeover of the cognitive work and problem solving, so that assembly workers primarily perform the remaining manual tasks.Accordingly, qualitative studies stress the risk of systematic deskilling of workers when introducing IASs at work as a result of the takeover of cognitive tasks (Baethge-Kinsky, 2020; Blumberg & Kauffeld, 2020).Consequently, assembly workers need to have fewer skills, knowledge, and abilities to complete assembly processes, particularly cognitive skills (Walczok & Bipp, 2023).On the other hand, the instruction function of the IAS could also increase the need for new skills, such as digital competencies to effectively use IASs.Building up on a framework model of digital competencies of employees (Oberländer et al., 2020), these could refer to the evaluation of digitalized information in the form of projected videos and basic knowledge about how the IAS works to handle potential false positive or false negative feedback on individual assembly steps (Oberländer et al., 2020) "alternation between tasks within a job that can require different skills and responsibilities but is not associated with a change to a different function or department" (Mlekus & Maier, 2021, p. 2).IASs are not only designed to support assembly workers at high task rotation but also several scholars postulate that task rotation could counteract negative impacts on work design and motivation, such as monotony and boredom that result from technologies taking over cognitive tasks (e.g., Mlekus et al., 2022).Hence, task rotation could represent a vital organizational implementation factor that affects the impact of technological changes in the workplace on work design (Gagné et al., 2022;Mlekus et al., 2022).
In their meta-analysis, Mlekus and Maier (2021) found a positive relationship between task rotation and attitudinal work outcomes (e.g., job satisfaction) which they suggest to be due to modified MWCs.They postulate that task rotation not only increases task variety per definition but also skill variety as the execution of numerous tasks requires a higher number of skills compared with jobs in which fewer tasks are executed (Mlekus & Maier, 2021).The proposed effects of task rotation on task and skill variety and work outcomes were supported by findings by Mlekus and Maier (2021).In two studies, they presented a hypothetical assembly workplace with a digital assistance system (which strongly resembles the IAS we focus on in its functions) to participants.The authors manipulated task rotation (no task rotation vs. task rotation) in a between-subjects design.In both experiments, task rotation elevated task and skill variety which in turn positively influenced work outcomes.
In the case of task rotation in assembly, we also anticipate positive effects of task rotation (alone and in combination with IASs) on MWCs.When working with task rotation, assembly workers execute a higher number of diverse assembly products resulting in a higher perceived task variety compared with working without task rotation.Hence, the successful assembly of diverse products requires a higher skill variety, primarily manual skills by manually operating with different assembly parts, materials, and tools, and learning numerous assembly processes.
H6. Skill variety is significantly higher when working with task rotation than when working without task rotation.
In addition to the impact of task rotation also postulated by Mlekus and Maier (2021), we propose that task rotation in assembly furthermore leads to higher information processing, as assembly workers learn and perform more diverse assembly processes in their daily work, and thus have to process more information than when no task rotation is implemented.
H7. Information processing is significantly higher when working with task rotation than when working without task rotation.
Since we anticipate that working with IASs and task rotation leads to enhanced information processing compared with working without IASs or task rotation, respectively, we postulate that a combination of both, working with IASs with task rotation, maximizes the amount of information that assembly workers need to process.This requires the processing of numerous, varying instructions that are projected by the IAS, and using varying tools, materials, mounting parts, and products.
H8.A combination of work with IASs and task rotation leads to higher information processing compared with other working conditions.
Given that we expect contradictory effects of the IASs on skill variety (cf.competing H5a and H5b), we investigate how the combination of IASs and task rotation impacts skill variety exploratively in a research question.On the one hand, an enhanced variability in assembly products by frequent task rotation could promote, for example, manual skills which could in turn counteract the reduction of required cognitive skills.On the other hand, both working with IASs and task rotation could boost digital competencies and manual skills, respectively.
RQ: How does the combination of work with IAS and task rotation impact skill variety?
Additionally, Mlekus and Maier (2021) postulate that task rotation will decrease work scheduling autonomy if "employees might be required to follow a fixed rotation roster" (p.3), and will increase task identity if task rotation leads to the execution of subsequent tasks that result in a holistic work process.However, we do not postulate an effect of task rotation on work scheduling autonomy or task identity.First, neither assembly workers who need to follow a fixed rotation schedule nor workers who continuously assemble identical products (no task rotation) can choose a desired order of tasks or products.Second, task rotation does not necessarily lead to the assembly of numerous assembly products that are further processed into a final overall product.
Third, we do not anticipate that the IAS affects task identity as the assembly work as a whole remains largely unchanged (Baethge-Kinsky, 2020).

| STUDY 1: WORKING WITH IAS IN A COMPLEX ASSEMBLY PROCESS
To test H1-H5 and the transferability of the effect of the IAS on MWCs by Walczok and Bipp (2023), we apply an experimental vignette methodology (EVM) study with two experimental conditions (work without IAS vs. work with IAS) and used a more complex assembly process which is also used in practice.This assembly process (twist stop) is characterized by a higher number (20) and more diverse assembly steps (Keller et al., 2019) which require more components and tools than the assembly of the simple box that was used in Walczok and Bipp (2023).Thereby, we postulate higher mental work, resulting in a higher degree of difficulty (Radowski et al., 2015).

| Experimental design and procedure
We applied an online EVM (between-subject) study design in resemblance to Walczok and Bipp (2023)   were informed that the product to be assembled would change every 2 h.Finally, participants rated the presented assembly workplace from their condition in terms of equipment use, and MWCs, and provided demographic information.This procedure builds up on the notion that job analysis is possible through observation (Dierdorff & Wilson, 2003).

| Participants
We invited German blue-collar workers, employees with prior work experience in manufacturing, and novices (workers from other fields and without prior work experience in manufacturing) from our professional and personal networks to participate in our online experiment.As IASs should assist people with low skills in training phases, we included novices as a specific target group of IASs (Doolani et al., 2020).We donated 0.50€ per participant to a charitable organization.

| Manipulation check
To ensure that we successfully manipulated the assembly workstation, the participants rated the equipment use using a validated German version (Stegmann et al., 2010) of the WDQ (Morgeson & Humphrey, 2006) on a five-point Likert scale ranging from strongly disagree (1) to strongly agree (5).We adapted the items for the manipulation check and MWCs ratings by switching from "my job" to "the job" in both studies to refer to the presented work situation.

| Measures
We measured MWCs using the validated German version (Stegmann et al., 2010) of the WDQ (Morgeson & Humphrey, 2006), again on a five-point Likert scale (Table 1).Internal consistency reliabilities varied from .66 (specialization) to .87 (feedback from job).In Study 1, we investigated whether we can transfer prior findings on the effect of IASs on MWCs by Walczok and Bipp (2023) to a more complex assembly process.In line with prior findings, we identified moderately higher feedback from job and weakly higher information processing when working with IASs compared with working without IASs.In contrast to prior findings, our current study provides evidence for restricted autonomy (in all three investigated autonomy facets) when working with IASs to a moderate degree.

| Results and discussion Study 1
Consequently, the results of our first study suggest that the implementation of IASs in assembly represents a double-edged sword in terms of MWCs.On the one hand, assembly workers benefit from the increased feedback from job and information processing as long as higher information processing does not lead to information overload (Walczok & Bipp, 2023).On the other hand, the restricted autonomy facets could counteract these two positive effects of IASs on MWCs, hence, questioning the added value of IASs.Therefore, our study is the first to provide causal evidence on the contradictory effects of IASs on MWCs.We partially transferred prior findings to a more complex assembly process, further highlighting the importance of their degree of difficulty for the evaluation of IASs and resulting effects on MWCs.
Since we did not find any differences in the remaining knowledge characteristics (job complexity, problem solving, specialization, and skill variety) between both experimental conditions, our results reinforce prior findings that the investigated IAS seems to fail to cognitively assist assembly workers (Walczok & Bipp, 2023).Thus, reductions in build times of small-scale assembly tasks depending on the number of performed assembly processes with the beginning of stagnation after only three assembly runs.This suggests that cognitive assistance in assembly tasks is specifically needed in highly frequent product changes.More rapid product changes (higher task rotation) could therefore be a prerequisite for IASs to be considered as cognitive relief.Therefore, we experimentally manipulated task rotation (in combination with the use of the IAS) in our second study.

| STUDY 2: THE ROLE OF TASK ROTATION
In our second EVM study with German employees, we investigated the effect of IASs, task rotation, and their interaction on MWCs (H1-H8, RQ) to further stress the importance of the highly dynamic product changes caused by individual customer requests.We preregistered the hypotheses and procedure before data collection and uploaded the used material to the Open Science Framework (https://osf.io/kteqy?view_only=44ac9c4d661348c9b8902312f4a2036b).

| Experimental design and procedure
We applied a 2 × 2 EVM (between-subject) study design with manipulation of the assembly workplace (work without IAS vs. work with IAS) and the task rotation (no task rotation vs. task rotation), resulting in four experimental conditions.We extended the study design from Study 1 and used the same instructions and baseline information.We presented participants from all experimental conditions with the hypothetical assembly workstation and the assembly of the twist stop.Besides the use of IAS (the assembly process was or was not supported by the IAS), task rotation was experimentally manipulated.Participants in the high task rotation conditions were informed that after every hour they would switch with a colleague to a nearby assembly station to carry out individual customer requests.After another hour, they switch again to assemble a different product.To amplify the illustration, we presented participants in the task rotation conditions a new assembly workstation with(out) IAS as well as the assembly of a box with(out) the assistance of the IAS from Walczok and Bipp (2023).

| Participants
As in Study 1, we invited German employees from professional and personal networks and used a donation of 0.75€ per participant to a charitable organization as an incentive.In total, 229 participants completed the experiment from December 2022 to February 2023.
We excluded three participants with technical issues during the presentation of the hypothetical workplace, five participants who failed the built-in attention checks, two participants with missing information in demographics, and nine who reported "student" as their current job from the analysis.(M = 38.97,SD = 13.57).Most participants were male (51.7%).Only 32 participants had prior work experience with IASs (18.2%).

| Manipulation checks
To ensure that the experimental manipulations of the assembly workstation and task rotation were successful, participants rated the equipment use and task variety, respectively, with the validated German version (Stegmann et al., 2010) of the WDQ (Morgeson & Humphrey, 2006).Comparing the ratings across the conditions supported that our manipulation was successful.First, equipment use (α = .70)was significantly higher in the conditions with IASs (M = 2.14, SD = 0.79) than without IASs (M = 1.84,SD = 0.72), t(174) = −2.689,p = .008,d = −0.407.

| Results and discussion Study 2
We conducted a series of ANOVAs to test our hypotheses (Table 4).In line with H1, we identified significantly higher feedback Moreover, surprisingly, the rotation of assembly products after every hour did not lead to a significant increase in skill variety as postulated by Mlekus and Maier (2021) or information processing and did not enhance the positive effect of IASs on information processing.Thereby, our results suggest that task rotation has limited potential to valorize work design in assembly or counteract the negative effects of technologies on MWCs, besides simply enhancing task variety.
Finally, the MWCs ratings in all four experimental conditions were low except for feedback from job.In particular, skill variety and information processing were also low, even in conditions with task

| GENERAL DISCUSSION
Applying EVM study designs across two German samples, we tested the effect of the IAS on MWCs in assembly by taking technological design factors and task rotation into account as an organizational implementation factor (Figure 1) based on the suggested model by Gagné et al. (2022).In detail, we provided causal evidence on the contradictory effects of IASs regarding MWCs by examining whether prior findings on the impact of IASs on MWCs in assembly (Walczok & Bipp, 2023) are transferable to more complex assembly processes and depend on the extent of task rotation.Our results fully replicate across our two studies: We found causal evidence for the effect of IASs in terms of increased feedback from job and information processing as well as restricted work scheduling, decision-making, and work methods autonomy by working with IASs in both of our studies.Therefore, our results imply that working with IASs represents a double-edged sword regarding MWCs.On the one hand, assembly workers receive more immediate feedback throughout the assembly process, which also increases the amount of information that needs to be perceived and processed and satisfies the need for competence as outlined in self-determination theory (Deci & Ryan, 1985), thereby increasing motivation.On the other hand, higher standardization of work processes by instructing workers with context-sensitive in situ projections results in decreased different facets of autonomy, hampering the satisfaction of the need for autonomy, and thus motivation (Gagné et al., 2022).
As Walczok and Bipp (2023) only identified enhanced feedback from job and information processing when executing a simple assembly task with the support of IASs, our results obtained with a more complex assembly process emphasize how the perception of IASs depends on the difficulty of the supported assembly tasks.If IASs are used for support in the execution of more objectively difficult assembly processes, contradictory effects on motivation were visible in our current study.Additionally, our results imply that the extent of task rotation is negligible when it comes to the perception of IASs and their impact on MWCs in assembly.Consequently, task rotation does not represent an adequate work design intervention to valorize MWCs in assembly or to boost the positive effects of technologiesin our case information processing.By this, our causal evidence contributes to the postulation of work design theories that highlight the role of technological changes in the workplace (Gagné et al., 2022) and provides much-needed practical implications for the implementation of IASs in modern assembly.Abbreviations: ANOVA, analysis of variance; IAS, intelligent assistance system.Gagné et al. (2022) seems to be too general to test how technologies change the way work is perceived and executed.We successfully identified increased feedback by working with IASs as it provides immediate feedback to individual assembly steps.Respectively, we found significantly enhanced information processing since workers need to perceive and process context-sensitive in situ projections in addition to traditional assembly work.Further and in line with qualitative studies (e.g., Blumberg & Kauffeld, 2020), we found restricted autonomy in all facets by working with IASs, resulting from the higher standardization of assembly processes and specification of assembly steps.However, since the IAS could potentially take over cognitive processes, we anticipated decreases in knowledge characteristics, namely, job complexity, problem solving, specialization, and skill variety.This implies that assembly workers gain similar knowledge, skills, and abilities by learning on the job despite the cognitive assistance from the IAS.Thus, our results contradict the anticipated systematic deskilling by the use of IASs which is crucial considering that the majority of organizational learning (estimations range from 70% to 90%) occurs informally through learning on the job (Cerasoli et al., 2018).Novel theories that acknowledge the such as reduced IAS acceptance and use if its output was perceived as complex (Javdan et al., 2023).

| Theoretical implications
Although

| Limitations and future research
Although the results of our two studies provide implications for the use of IASs and work design theory, we have to acknowledge several limitations.First, while EVM study designs have the potential to maximize internal and external validity, thereby providing causality and high generalizability of findings (Aguinis & Bradley, 2014), studies examining how accurately our results translate to a real-life setting need to be conducted.Therefore, we recommend the investigation of work design in an assembly setting before and after the IAS is implemented, and its consequences for attitudinal work outcomes, such as work motivation or job satisfaction.This would also allow investigation of how technological interferences of the system that affect productivity and accuracy (Bortolini et al., 2021) influence the perception of IASs (Walczok & Bipp, 2023).Furthermore, future studies in a real production setting might investigate whether and how employees proactively react to adapt work design after the implementation of innovative technologies (Gagné et al., 2022).
Second, despite presenting a more complex assembly process, MWCs ratings were low across all experimental conditions in both studies.Even more difficult assembly processes should be used in future studies to further counteract floor effects in MWCs ratings.
However, these did not prevent us from identifying reductions in the already low autonomy (in the work without IASs) by the implementation of IASs.The restricting effect of IAS on autonomy could be even stronger in magnitude in more difficult assembly processes that also external validity, we aimed to maximize the level of immersion by presenting the hypothetical situation with text, picture, and video material.Participants were randomly assigned to one of the two conditions.First, all participants received a short text about a hypothetical situation.They were instructed to imagine themselves working as assembly workers in the production halls of an assembly company.Next, we presented a hypothetical assembly workplace using a short text and a picture with IAS for participants in the work with IAS condition and without IAS for participants in the work without IAS condition.Additionally, participants in the IAS condition received information on the functions of the IAS.Then, all participants watched a short video of the assembly of a twist stop, again with or without the support of the IAS, according to their experimental condition.Participants in both experimental conditions two experimental conditions work with IASs and work without IASs across all nine MWCs, F(9, 159) = 5.305, p < .001,η p 2 = 0.231.One-way MANCOVA with expertise level (prior or current work experiences in manufacturing vs. no experiences in manufacturing) as covariate yielded identical results, F(9, 158) = 5.288, p < .001,η p 2 = .231,as experts and novices did not significantly differ in their ratings of MWCs according to their experimental condition, F(9, 158) = 0.849, p = .572,η p 2 = 0.046.We conducted subsequent t-tests for the MWCs separately to test our hypotheses.We found significantly higher feedback from job in work with IASs than in work without IASs, t(167) = −3.709,p < .001,d = −0.573,supporting H1.Information processing was also significantly higher in work with IASs than in work without IASs, t(167) = −1.984,p = .049,d = −0.307,supporting H2.We found significantly lower work scheduling autonomy, t(146.69)= 3.347, p = .001,d = 0.527, decision-making autonomy, t(100.15)= 3.887, p < .001,d = 0.643, and work methods autonomy, t(124.98)= 3.346, p = .001,d = 0.538, in work with IASs than in work without IASs, supporting H3a-c.We did not identify any significant differences between work with IASs and work without IASs in job complexity, t(167) = −0.127,p = .899,d = −0.020,problem solving, t(167) = 0.270, p = .787,d = 0.042, or specialization, t(167) = 0.382, p = .703,T A B L E 2 Reliabilities, cell means, and standard deviations in motivational work characteristic ratings in work without IAS and work with IAS (Study 1).
the absence of cognitive relief despite the in situ step-by-step instructions by the IAS remains.To achieve this desired positive effect of the technological design factors on work design, we need to take organizational implementation factors and their interaction with technological design factors into consideration as outlined in the model by Gagné et al. (2022).As IASs are mainly used for learning new assembly processes (Egger-Lampl et al., 2019), which frequently change due to individual customer requests, high task rotation could play a fundamental role in the beneficial role of IASs with regard to work motivation.Although in the current study, all participants received the information that the task would change after 2 h, this description of potential task rotation might have been too abstract or discreet, or the used time interval of 2 h to the next rotation might be too long for the cognitive relief by the IAS.This is also in line with research by Watson et al. (2010) who demonstrated drastic from job in work with IASs than in work without IASs, F(1, 172) = 39.216,p < .001,η p ² = 0.186.We also found significantly higher information processing in work with IASs than in work without IASs, F(1, 172) = 4.345, p = .039,η p ² = 0.025, supporting H2.In congruence with H3a-c, our results indicate significantly lower work scheduling autonomy, F(1, 172) = 12.734, p < .001,η p ² = 0.069, decision-making autonomy, F(1, 172) = 5.210, p = .024,η p ² = 0.029, and work methods autonomy, F(1, 172) = 5.878, p = .016,η p ² = 0.033, in work with IASs than in work without IASs.We did not find any significant differences between work with IASs and work without IASs in job complexity, F(1, 172) = 0.915, p = .340,η p ² = 0.005, problem solving, F(1, 172) = 0.810, p = .369,η p ² = 0.005, or specialization, F(1, 172) = 1.959, p = .163,η p ² = 0.011, providing no evidence for H4a-c.We also found no significant difference in skill variety between work with IASs and work without IASs, F(1, 172) = 0.809, p = .370,η p ² = 0.005, rejecting H5a and H5b.In terms of task rotation, we did not identify significant main effects on skill variety, F(1, 172) = 0.007, p = .931,η p ² = 0.000, or information processing, F(1, 172) = 0.102, p = .750,η p ² = 0.001, rejecting H6 and H7, respectively.Contrary to H8, we found no significant interaction effect of IASs and task rotation on information processing, F(1, 172) = 0.001, p = .976,η p ² = 0.000.Finally, with regard to our RQ, we did not identify a significant interaction effect of IASs and task rotation on skill variety, F(1, 172) = 0.128, p = .721,η p ² = 0.001.Besides examining the impact of IASs on work design, we investigated how high task rotation as an organizational implementation factor interacts with the effect of IASs on MWCs in assembly in Study 2. As we found the main effects of IASs in terms of increased feedback from job and information processing as well as restricted work scheduling, decision-making, and work methods autonomy in working with IASs (compared with working without IAS), the results of Study 2 reinforce the notion of IASs as a double-edged sword regarding MWCs.Whereas the largely increased feedback from job and the weakly enhanced information processing have a motivationenhancing effect, the moderately restricted work scheduling autonomy as well as weakly reduced decision-making and work methods autonomy counteract these positive motivational effects.Again, the IAS did not affect other knowledge characteristics, therefore not changing requirements in skills, abilities, and knowledge that assembly workers need to successfully perform assembly work daily.Hence, the results of Study 2 demonstrate that the investigated IAS does not lead to the intended cognitive relief.This is further stressed by the fact that knowledge characteristics are not affected by the implementation of task rotation or IAS with high task rotation in combination.Thus, our results refute the potential benefits (enhancing skill variety by promoting both manual skills and digital competencies) of IASs in the cognitive support of assembly workers in highly frequent product changes caused by individual customer requests.

F
I G U R E 3 Motivational work characteristic ratings according to the experimental conditions (Study 2).IAS, intelligent assistance system; MWCs, motivational work characteristics.Flagged variables represent the main effects of the IAS.Error bars display standard deviations.*p < .05;***p < .001.WALCZOK and BIPP | 215 rotation.This study stresses the importance of work design interventions other than task rotation in assembly to improve MWCs, and hence, the motivation of assembly workers.
impact of technologies on work design should consider which skills remain relevant and which skills are negligible in work with innovative technologies due to automation.These should integrate existing frameworks such as the cyber-physical systems transformation framework by Waschull et al. (2020) which highlights the contradictory effects of innovative technologies on the substitution and creation of tasks in light of automation, thereby leading to enriched, simplified, and substituted jobs.Additionally, we did not find any evidence that task rotation alters investigated MWCs besides task variety or interacts with technological design factors of the IAS.This emphasizes the requirement to specify in detail which technological design and organizational implementation factors (and both in combination) modify work design.Since the terms technological design and organizational implementation factors remain rather vague and the model by Gagné et al. (2022) neglects further determinations of which functions of technologies have specific effects on work design, the results of our two studies contribute to the further specification of this framework.By demonstrating that task rotation as an organizational implementation factor does neither improve MWCs nor boost the positive effects of IASs on MWCs, we highlight the need to extend the postulated framework.Gagné et al. (2022) state "[h]ighly skilled individuals or those with proactive personalities might actively shape the technology and/or craft their work design to better meet their needs and increase their motivation" (p.383).By statistically controlling for the participants' expertise level, we have first indications that the skill level in terms of prior experience in manufacturing does not play a fundamental role in how IASs are perceived.Thus, our results stress the importance of specific theories and models that allow one to predict and test how new technologies (e.g., IASs) shape MWCs, and that human-centered design and potential effects on MWCs should be considered during the development of technologies.Our findings show that besides task variety, MWCs remained unchanged by the implementation of task rotation challenges the notion that IASs are developed to intelligently support workers in highly dynamic product changes and questions the benefit of this work design intervention.Missing and small effects of task rotation in the meta-analysis by Mlekus and Maier (2021) on attitudinal work, learning, and psychological health outcomes could be attributed to the fact that task rotation has weaker impacts on MWCs than expected, especially on skill variety.4.2 | Practical implicationsOur studies provide practical implications for the implementation of IASs in assembly.Contrary to the results byWalczok and Bipp (2023) who emphasize a motivation-enhancing effect of IAS by increasing feedback from job and information processing, our two studies show contradictory results.As indicated in qualitative studies (Baethge-Kinsky, 2020; Blumberg & Kauffeld, 2020), we identified restricted autonomy, specifically work scheduling, decision-making, and work methods autonomy by working with IASs.Practitioners should therefore not expect purely positive effects of IASs on MWCs in assembly and ensure that workers perceive autonomy in other ways, for example, through flexible shift starts.To avoid the negative longterm effects of IASs on autonomy, practitioners should consider selfadaptive IAS(Yigitbas et al., 2023) that reduces the level of support with increasing skill level.Although we identified reduced autonomy in all facets when working with the investigated self-adaptive IAS, the option of high self-adaption may nevertheless represent a promising solution to counteract negative effects on autonomy.By reducing the projection of instructions with increasing execution of identical assembly products and thus gaining skills, the self-adaption of IASs enables the execution of assembly steps in preferred ways.However, this could counteract the positive effects of enhanced feedback from job and information processing as these decrease with reduced support of the IAS, too.Replicating prior findings(Walczok & Bipp, 2023), the IAS does not impact further knowledge characteristics job complexity, problem solving, specialization, and skill variety, our results refute the risk of systematic deskilling of employees (Baethge-Kinsky, 2020;Blumberg & Kauffeld, 2020) through cognitive support by the IAS.Nevertheless, this also indicates that cognitive support as its primary goal is not achieved by the investigated cognitive-assistive IAS.Practitioners should consider the missing decreases in knowledge characteristics when implementing IASs in assembly.Given that IASs should have the potential to reinforce the inclusion of workers with cognitive deficits(Mark et al., 2019), our findings imply that IASs are not suitable for doing so.Particularly, increased information processing could result in information overload(Walczok & Bipp, 2023).Roetzel (2019) argues that information overload frequently results from the misuse of ICT technologies at work which leads to decreased decision-making performance, resulting in hampered performance when working with IASs.A higher amount of information could pose further problems, WALCZOK and BIPP | 217 provide more autonomy.Additionally, subsequent studies should focus on how self-adaptive IASs that reduce the level of support based on workers' skill level provides a solution to counteract restricted autonomy.Third, although we measured MWCs with a validated German version of the WDQ(Stegmann et al., 2010), internal consistency reliabilities were low for work scheduling autonomy in Study 1, and specialization in both studies.Hence, results regarding work scheduling autonomy and specialization should be interpreted with caution.Fourth, the generalizability of the effect of the investigated IAS on MWCs to other IASs remains an empirical question.Given that we based our hypotheses on the functions of the specific IAS in question, the generalizability of the results may be limited to other cognitiveassistive IASs with a low level of support that are characterized by similar functions (providing feedback for subsequent assembly steps and instructing assembly workers with sensitive in situ material).These might include pick-to-light systems or projection-and augmentationbased IASs(Walczok & Bipp, 2023).Finally, we encourage the experimental manipulation of technological design factors (such as the extent of feedback and in situ projections) to attribute specific effects to specific functions of technologies.5 | CONCLUSIONBy identifying enhanced feedback from job and information processing, as well as restricted work scheduling, decision-making, and work methods autonomy by working with a cognitive-assistive IAS in assembly, our two studies provide vital causal evidence of their contradictory effects regarding motivational work design.Our results suggest no effect of the IAS on the knowledge characteristics job complexity, problem solving, specialization, and skill variety.Despite being developed to support assembly workers in highly dynamic product changes due to individual customer requests, the effect of IASs on MWCs does not interact with the extent of task rotation.Our results stress the importance of technology for work design and highlight the need for further refinement of work design theories to make accurate predictions of positive and/or negative effects.

Table 2 and
Figure 2 show the descriptive results of the two experimental conditions.In general, on a descriptive level, low ratings in all MWCs-except feedback from job-are visible.One-way multivariate analysis of variance (MANOVA) indicated a significant 1 They answered the item "How easy was it for you to put yourself in the situation presented" with very difficult (1), difficult (2), or rather difficult.
Notes: Cronbach's alpha is in parentheses in the first column.
IASs are developed to support workers cognitively in highly dynamic product changes, our results suggest that the extent of task rotation does not affect the contradictory effects of IASs regarding MWCs.Consequently, our results suggest that IASs are not an adequate technology to cognitively assist workers in highly dynamic product changes due to individual customer requests.Equally, our results indicate that task rotation does not represent an appropriate work design intervention to improve work design in assembly besides promoting task variety.Thus, also task rotation and working with IASs do not lead to an overall increased skill variety by boosting manual skills and digital competencies.Practitioners who aim to improve MWCs should therefore rely on other work design interventions.Employees can also proactively alter MWCs after the implementation of new technologies using job crafting as outlined by Gagné et al. (2022).