An HCI Study on Trust, Explainability, and Player Analytics Interfaces
AI-powered analytics are now widely used in football scouting and recruitment, yet little is known about how their outputs are actually interpreted and trusted. This study looks at the human side of those systems.
How do football-literate stakeholders interpret and make sense of AI-driven player performance metrics through interaction with a scouting and recruitment analytics interface?
How do specific interface design elements, particularly radar visualisations and AI-generated narrative summaries, influence how people interpret player performance metrics?
How is trust in an AI-supported analytics interface formed, negotiated or resisted during interaction?
Twelve Football, an AI-powered football intelligence platform, with a focus on its Earpiece feature: AI-written summaries of a player's strengths and weaknesses, set against comparable players.
10 football-literate stakeholders, including coaches, athletes and analysts, each using the tool on real player profiles.
Remote sessions of 45 to 60 minutes combining the think-aloud protocol with semi-structured interviews.
Qualitative thematic analysis within an interpretivist HCI approach, focused on how meaning and trust are built, not on model accuracy.
Even with clear summaries and visuals, users often felt they could not see why the AI reached its conclusions.
Participants checked every AI claim against their own football knowledge. Domain expertise mattered more than interface quality.
Trust was situational and negotiated. The system was treated as an instrument that supports judgment, not a decision maker.
The AI should be an instrument for expert judgment, not a replacement for it.
This thesis brings my principles together: agency over automation, calibrated trust, and clarity at the point of action, tested with the people who make real decisions.
Curious about the details? Ask the assistant in the corner, it has read the full thesis.