AI Reliance in the Fantasy American Football Lineup-Setting Task

Authors:
Open Access
Publisher website
Google Scholar
Abstract
Analysts, athletes and coaches have long relied upon decision support tools, but new AI capabilities raise questions of whether they can “appropriately rely” on these models. In this work, we explore the starting lineup decisions of fantasy American football managers in the presence of projected points. Across more than 2.5 million lineup-weeks from a single fantasy platform during the 2025 NFL season, we find that fantasy managers’ starting lineups appropriately rely on AI advice in only 15.8% to 26.8% of cases. Despite managers’ observed near complete reliance, limited deviation from AI advice is often harmful to their decision-making. Bridging literature from critical data studies and AI reliance, we discuss possible explanations for this behavior, and offer design directions towards complementarity in this setting. Finally, we highlight tensions revealed by the case of fantasy football, and invite researchers to consider SportsHCI as a site to bridge social desirability and technical possibility in and beyond sports.
Funding Information
  • National Science Foundation (2313998)

This publication has 73 references indexed in Scilit: