Playtech joins Codex Talk alongside ARM, Darktrace and NVIDIA to debate The Need for Explainability in AI

Playtech discusses the need to harness explainability when applying AI to sensitive domains.

Categories

Simo Dragicevic discusses the importance of explainability in AI systems used for consumer protection in gambling. He outlines the challenges of machine learning in sensitive domains and presents a case study on how BetBuddy integrates explainability to meet regulatory and ethical standards.

Key Insights:

• Industry Context:
  Gambling is a €400B+ global industry facing increasing scrutiny, especially in markets like the UK and Italy. Regulatory backlash includes advertising bans and proposed stake limits, raising the need for smarter, responsible innovation.
• AI vs Human Judgment:
  AI excels at “reckoning” (automation) but lacks “judgment”—ethical reasoning and contextual awareness. In sensitive areas like gambling harm detection, trust and transparency are essential.
• Accuracy vs Explainability:
  Stakeholders prefer less accurate but explainable models over black-box systems. This reflects discomfort with opaque decision-making in high-impact areas.
• Challenge of Machine Learning:
  Unlike traditional software, ML models are shaped by data, making them hard to verify. Thousands of nonlinear connections mean decisions can’t be easily traced or debugged.
• Solution – Explainable AI Layer:
  BetBuddy developed a system that interprets model outputs, showing which behavioral indicators (e.g. deposit increases, session length) drive risk predictions. This allows for validation, debugging, and stakeholder trust.
• Applications in Player Interaction:
  Explainability enables targeted, personalized messaging. For example, players increasing deposits can be nudged toward setting limits, or those showing signs of fatigue can be encouraged to take breaks—without labeling them as “problem gamblers.”
• Ethical and Strategic Value:
  Explainable AI not only meets regulatory requirements but improves customer experience. It supports human-centric design, ongoing evaluation, and transparency about blind spots.

As AI becomes more pervasive, the next decade must focus on ethics, explainability, and evaluation. Without these, industries risk losing trust and failing to fully realize AI’s potential in consumer protection.