The Social-Proof Effect in AI and Expert Framing: Experimental Evidence from Football Outcome Predictions

Abstract

This study investigates how framing conditions influence predictive judgments in professional football. A survey experiment was conducted with 290 respondents recruited via Prolific, each of whom evaluated real fixtures from Brazil’s Série B. For every match vignette, participants were randomly exposed to one of four conditions: a neutral presentation, betting odds emphasis, expert attribution, or AI attribution. The design generated multiple decisions per respondent, enabling econometric analysis with respondent-clustered standard errors. Results show that framing significantly shaped predictions. The odds presentations reinforced market expectations, while AI framing consistently reduced the likelihood of choosing the favorite, indicating strong algorithmic influence as they were contradicting the betting odds. By contrast, expert framing had weaker and less consistent effects. Stepwise logit models, extended with the match uncertainty measured by the odds difference, and robustness checks using alternative estimators confirm the reliability of these findings. The study extends research on social proof, framing, and algorithm appreciation by showing how credibility cues operate in sports contexts. It also highlights the potential risks of algorithmic persuasion in betting and fan engagement, with broader implications for how information is presented in sport management and digital consumption environments.