When AI Delivers Earnings: Experimental Dataset on Digital Spokespersons, AI-Authored Financial Disclosure, Trust, and Investment Allocation (2×2 Investor Experiment)
Description
This dataset contains participant-level responses from a randomized 2×2 behavioral experiment examining how AI-mediated financial communication affects investor perceptions and investment decisions. The experiment investigates how communicator type (human executive vs. digital avatar) and disclosure of content authorship (management-authored and compliance-reviewed vs. AI-authored with limited human review) jointly influence perceived accountability, trust in financial communication, and capital allocation behavior. Participants viewed a standardized approximately 75-second earnings update video for a simulated publicly listed company, Aurevia Technologies Inc. The financial content, script, and visual information presented in the video were held constant across experimental conditions. Only the communication interface (human executive or digital avatar) and the disclosure regarding the origin of the message content varied across participants. After viewing the video, participants completed a structured questionnaire measuring manipulation checks, perceptions of accountability and governance, trust and credibility toward the earnings communication, and an incentivized-style allocation decision in which respondents allocated a hypothetical $10,000 investment between Aurevia stock and a risk-free asset. The dataset includes demographic variables, investment background information, risk tolerance measures, baseline attitudes toward AI in financial decision contexts, and multiple Likert-scale measures of trust, credibility, and accountability perceptions. The primary behavioral outcome variable records the amount allocated to Aurevia stock. Additional variables include manipulation-check indicators and exclusion flags used for robustness and data-cleaning procedures. The data support reproducible statistical analyses of the main and interaction effects of AI-mediated communication and AI-authored financial disclosure on investor trust and allocation behavior.
Files
Steps to reproduce
The dataset originates from a randomized 2×2 online behavioral experiment designed to examine how AI-mediated financial communication influences investor trust and investment decisions. Participants were randomly assigned to one of four experimental conditions defined by two factors: (1) communicator type (human executive vs. digital avatar) and (2) disclosure of content authorship (management-authored and compliance-reviewed vs. AI-authored with limited human review). All participants viewed a standardized earnings-update video (~75 seconds) for a simulated publicly listed company, Aurevia Technologies Inc. The financial information, script, and visual slides were identical across all conditions. Only the communication interface (human presenter vs. digital avatar) and the disclosure regarding the origin of the message content varied between conditions. The experimental procedure followed these steps: Participants accessed an online survey environment and first provided basic demographic and investment background information (age, investment experience, trading frequency, and financial knowledge screening). Participants were randomly assigned to one of the four experimental conditions. Before the video, participants viewed a short disclosure statement describing either management-authored/compliance-reviewed content or AI-generated content with limited human review. Participants then watched the earnings update video corresponding to their assigned condition. Immediately after the video, participants completed manipulation-check questions to verify recognition of the presenter type and the authorship disclosure. Participants then responded to survey items measuring perceived accountability, governance oversight, trust, credibility, transparency, and communicator competence using Likert-scale responses. Participants completed comprehension and recall questions about the financial information presented in the video. Participants made a behavioral investment decision by allocating a hypothetical $10,000 between Aurevia stock and a risk-free asset. Additional control variables were collected, including financial risk tolerance and baseline trust in AI systems. The dataset contains the raw responses for all variables, condition indicators, manipulation checks, and exclusion flags used to identify inattentive responses or failed comprehension checks. The dataset can be analyzed using standard statistical methods for experimental designs, including difference-in-means tests, regression models, and interaction analysis for the two experimental factors.
Institutions
- Ca' Foscari University of VeniceVeneto, Venice