When Some Use AI and Some Don’t : Felt Authorship, Asymmetric Attribution, and the Labour of Integration in AI-Assisted Group Projects
Nderemani, Moses (2026)
2026
Master's Programme in Sustainable Societies and Digitalisation
Informaatioteknologian ja viestinnän tiedekunta - Faculty of Information Technology and Communication Sciences
Hyväksymispäivämäärä
2026-06-02
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-202606026771
https://urn.fi/URN:NBN:fi:tuni-202606026771
Tiivistelmä
This thesis investigates how the use of generative AI changes the way people work together in groups. It covers three things: how much people feel ownership of AI-generated content, when they tell teammates that AI helped (or hide that it did), and how groups share credit and responsibility for AI-assisted work. Earlier studies have looked at each of these on their own, mostly with one person working alone with an AI tool. What happens when these dynamics meet in a group project, where teammates form the audience and where members use AI in different ways, has not been studied in a systematic way.
The study used a mixed-method online survey of university students and working professionals who had used AI in group projects. The survey combined scaled questions on ownership, disclosure, and credit with open questions that asked respondents to describe one specific moment in their own words. The final sample is 31 respondents with 29 valid responses, spanning light, shaped, and configured AI users. The scaled answers were summarised as simple counts and cross-tabulated against the respondent's position on an AI use spectrum. The open ended answers were analysed thematically.
Three findings stand out. First, how much the work felt like the user's own tracked how much of the visible contribution they had produced by hand, not how deeply they had set up the tool. This is the opposite of what the framework predicted. Second, credit and blame were uneven. Of the 29 respondents, 55% said they took personal responsibility when AI work went wrong, but only 14% claimed personal credit when AI work went well. Third, in groups whose members disagreed about whether to use AI at all, AI-using members produced visible output quickly while non-using members carried the load of reviewing, editing, and integrating that output. Disclosure rules alone could not move this work back across the group. The findings extend Pierce et al.'s theory of psychological ownership, suggests some bounds on Goffman's frontstage and backstage division by showing it applies most clearly when group norms are unset or contested, and place AI integration within Star and Strauss's account of how invisible work supports visible work. The thesis suggests two practical responses: tools that automatically track and mark AI-generated parts of a produced artifact, and group norms that take into account the work of integrating AI outputs as labour that deserves credit just like producing it does.
The study used a mixed-method online survey of university students and working professionals who had used AI in group projects. The survey combined scaled questions on ownership, disclosure, and credit with open questions that asked respondents to describe one specific moment in their own words. The final sample is 31 respondents with 29 valid responses, spanning light, shaped, and configured AI users. The scaled answers were summarised as simple counts and cross-tabulated against the respondent's position on an AI use spectrum. The open ended answers were analysed thematically.
Three findings stand out. First, how much the work felt like the user's own tracked how much of the visible contribution they had produced by hand, not how deeply they had set up the tool. This is the opposite of what the framework predicted. Second, credit and blame were uneven. Of the 29 respondents, 55% said they took personal responsibility when AI work went wrong, but only 14% claimed personal credit when AI work went well. Third, in groups whose members disagreed about whether to use AI at all, AI-using members produced visible output quickly while non-using members carried the load of reviewing, editing, and integrating that output. Disclosure rules alone could not move this work back across the group. The findings extend Pierce et al.'s theory of psychological ownership, suggests some bounds on Goffman's frontstage and backstage division by showing it applies most clearly when group norms are unset or contested, and place AI integration within Star and Strauss's account of how invisible work supports visible work. The thesis suggests two practical responses: tools that automatically track and mark AI-generated parts of a produced artifact, and group norms that take into account the work of integrating AI outputs as labour that deserves credit just like producing it does.