Work

Blueprint Preregistration: Interpersonal Transparency and Concealment of Generative AI Use (Birthday Messages to Parents)

Year: 2026
Type: preprint
Study design: Study Protocol
Abstract: The broader project consists of three interconnected studies: - Study 1a: Survey on generative AI descriptive and injunctive norms, disclosure thresholds, and disclosure and revelation judgments - Study 1b: Factorial vignette experiment on task, depth of generative AI involvement, and frequency of reliance - Study 2: Study of self-disclosure, transparency, concealment, and privacy-management practices in relation to generative AI use This preregistration covers study 2. For the other two studies, please see the broader project description this preregistration is embedded in. The study examines how people manage the interpersonal visibility of their generative AI use in a concrete relational communication context: producing a birthday message for a parent. It constitutes the first task-specific implementation of study 2 in a broader research program on social norms, disclosure, and evaluations of generative AI use. This preregistration describes the common design and analysis framework for the first phase of study 2. The initial study 2 phase focuses on five relational tasks: producing a birthday message for a parent; producing a condolence message for a close friend who has lost a close family member; producing a family trip itinerary to share with family members; producing a break-up message for a lover; and producing an email asking an acquaintance for a favour. The birthday-message-for-a-parent survey serves as the initial blueprint and was refined through pretesting before this registration. The same core What–Who–How–Why structure, conceptual distinctions, and principal measures will be adapted to the remaining tasks using task- and relationship-specific wording. Data collection may therefore take place sequentially across several task-specific surveys rather than in a single questionnaire. Participants are recontacted from two earlier studies in which they reported using generative AI in connection with producing a birthday message for a parent (1a, 1b). Recruitment takes place through Prolific. At the beginning of the survey, participants are asked to confirm that they have actually used generative AI for this task and that the resulting birthday message was shared with their parent. Only respondents who confirm both conditions proceed to the substantive survey. The study focuses on a real episode from each participant's own experience. The questionnaire follows a What–Who–How–Why structure. WHAT reconstructs the focal episode, including when it occurred, the generative AI tool used, the extent of AI involvement, and the respondent's familiarity with using generative AI for this type of task. WHO examines who, as far as the respondent knows, was aware that generative AI had been involved, distinguishing between the respondent, the parent who received the message, and relevant third parties. HOW examines how the AI use became visible or remained undisclosed, distinguishing practices such as proactive disclosure, disclosure after being asked, detection or accidental revelation, passive non-disclosure, strategic non-disclosure, strategic ambiguity, and active concealment. Where relevant, the survey also measures the extent and framing of disclosure and concrete practices used to make AI involvement less noticeable. WHY examines the episode-specific considerations that shaped how participants managed whether their parent knew about the generative AI use. These include perceived recipient entitlement and production-process privacy, authenticity and care, anticipated social evaluation, normalization and perceived relevance of AI use, and considerations related to honesty and possible later revelation. A short generalization module additionally assesses broader orientations toward the interpersonal transparency and privacy of generative AI use. An open-ended question invites participants to describe in their own words what makes generative AI use something that others should know about versus something they regard as private or not worth mentioning. Where possible, participants' responses will be linked to their earlier responses from studies 1a and 1b. This enables exploratory comparisons between previously measured judgments about generative AI use and disclosure and participants' subsequently reported real-world practices. Particular attention will be paid to the distinction between ordinary or passive non-disclosure and more deliberate forms of ambiguity or concealment. The study is primarily designed to provide a descriptive and explanatory account of how generative AI use is managed interpersonally in a relational context. Expected outcomes include estimates of how frequently AI use remains known only to the user versus becoming known to recipients or third parties; a characterization of recurring disclosure, non-disclosure, and concealment practices; and evidence about the considerations associated with different forms of visibility management. The study does not assume that non-disclosure is equivalent to concealment or deception. The questionnaire was refined following a small pretest conducted prior to this registration. Participants in that pretest will not be included in the preregistered main study (20 respondents, launched and completed on 12 August 2026). This registration concerns the birthday-message-for-a-parent task specifically. Subsequent Study 2 surveys may apply the same general framework to other interpersonal and relational uses of generative AI and will be documented separately.
Language: English
Open Access status: green
OpenAlex ID: W7203548857
Cites:
Cited by:
Related to: 10