Challenge and Hindrance Appraisals and Causal Attributions in AI-induced technostress : Critical Incident study of Knowledge Workers
Vornanen, Salla (2026)
2026
Tietojohtamisen DI-ohjelma - Master's Programme in Information and Knowledge Management
Johtamisen ja talouden tiedekunta - Faculty of Management and Business
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2026-06-08
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-202606087104
https://urn.fi/URN:NBN:fi:tuni-202606087104
Tiivistelmä
Generative artificial intelligence (AI) tools are being adopted into knowledge work rapidly and often without established practices to guide their use. Such tools can support and accelerate work, but they can also become a source of stress. This thesis examines AI-induced technostress as it is experienced and interpreted by knowledge workers themselves. It addresses two parallel questions: how knowledge workers appraise AI-related demands as challenge or hindrance stressors, and to which targets they attribute the causes of the outcomes of their AI-supported tasks when stress is present.
The study draws on semi-structured interviews with 21 knowledge workers in five Finnish companies, conducted using the Critical Incident Technique, in which participants described concrete positive and negative episodes of working with generative AI tools. Data was analyzed through Codebook Thematic Analysis. The analysis combined a transactional, appraisal-based view of stress and the challenge–hindrance distinction with a causal attribution lens based on the locus of causality. 28 incidents were analyzed for the appraisal question and 26 for the attribution question.
The analysis identified six AI-induced demands: output volume handling, managing AI use, input specification, output verification, iterative refinement, and competence building. Within these demands, thirteen hindrance stressors and eight challenge stressors were distinguished. Five of the six demands were appraised as both a challenge and a hindrance across incidents.
Hindrance appraisals were more frequent in the data. Knowledge workers attributed task outcomes to four targets: the user, the AI tool, the task, and the vendor behind the tool. The attributions were asymmetric across outcomes (success vs. fail). For failed outcomes, the most prevalent pattern was internal: workers tended to locate the cause in their own prompting, context-setting, or domain knowledge, absorbing the failure as their own. For successful outcomes, attributions surfaced less readily and were more diffuse, most often shared between the worker and the tool or assigned to the tool alone.
The findings indicate that the demands of generative AI are dual and appraisal-dependent, and that some of them are captured only partially by the established technostressor taxonomy. The prevailing tendency to internalize the cause of AI failures runs counter to common expectations from attribution research and may reflect the absence of shared practices for working with AI. For organizations, the results suggest that the productivity case for AI should be weighed against the verification, iteration, and effort it requires, that shared ways of working may reduce misplaced self-blame, and that the capacity to manage one's own use of AI is itself a competence worth supporting.
The study draws on semi-structured interviews with 21 knowledge workers in five Finnish companies, conducted using the Critical Incident Technique, in which participants described concrete positive and negative episodes of working with generative AI tools. Data was analyzed through Codebook Thematic Analysis. The analysis combined a transactional, appraisal-based view of stress and the challenge–hindrance distinction with a causal attribution lens based on the locus of causality. 28 incidents were analyzed for the appraisal question and 26 for the attribution question.
The analysis identified six AI-induced demands: output volume handling, managing AI use, input specification, output verification, iterative refinement, and competence building. Within these demands, thirteen hindrance stressors and eight challenge stressors were distinguished. Five of the six demands were appraised as both a challenge and a hindrance across incidents.
Hindrance appraisals were more frequent in the data. Knowledge workers attributed task outcomes to four targets: the user, the AI tool, the task, and the vendor behind the tool. The attributions were asymmetric across outcomes (success vs. fail). For failed outcomes, the most prevalent pattern was internal: workers tended to locate the cause in their own prompting, context-setting, or domain knowledge, absorbing the failure as their own. For successful outcomes, attributions surfaced less readily and were more diffuse, most often shared between the worker and the tool or assigned to the tool alone.
The findings indicate that the demands of generative AI are dual and appraisal-dependent, and that some of them are captured only partially by the established technostressor taxonomy. The prevailing tendency to internalize the cause of AI failures runs counter to common expectations from attribution research and may reflect the absence of shared practices for working with AI. For organizations, the results suggest that the productivity case for AI should be weighed against the verification, iteration, and effort it requires, that shared ways of working may reduce misplaced self-blame, and that the capacity to manage one's own use of AI is itself a competence worth supporting.