The Dark Side of AI in Insurance: A Systematic Review of Mechanisms Linking AI Design Features to Consumer Harm

The Dark Side of AI in Insurance: A Systematic Review of Mechanisms Linking AI Design Features to Consumer Harm

Summary

This systematic review (33 empirical studies) examines how AI design features in insurance — notably algorithmic opacity, hyper-personalisation and data-driven bias — lead to consumer harm. Using the TCCM (Theory–Context–Characteristic–Method) mapping the authors find theoretical fragmentation, geographic concentration of evidence and methodological imbalance in the literature. To offer a dynamic mechanism account, they propose a Trigger–Psychology–Decision–Outcome (TPDO) framework that traces how AI triggers consumer psychological responses which then shape decisions and harmful outcomes. Key psychological mediators identified include fairness concerns, anxiety and perceived loss of control; resulting behaviours include disengagement, resistance and exclusion, worsening trust and privacy outcomes. The review closes with a research agenda and governance suggestions to centre consumer protections in algorithmic insurance.

Key Points

• AI design features implicated: algorithmic opacity, extreme personalisation and data-driven bias.
• Psychological pathways matter — triggers (design features) → psychology (fairness concerns, anxiety, loss of control) → decisions → consumer harm.
• The authors synthesise evidence from 33 empirical studies and highlight theoretical and methodological fragmentation in the field.
• The new TPDO framework maps sequential mechanisms and helps move beyond static taxonomies of harms.
• Adverse outcomes documented include privacy erosion, service exclusion, reduced trust and consumer disengagement.
• Policy and research recommendations focus on consumer-centric governance, more geographically diverse studies and balanced methods (qualitative + quantitative).

Content summary

The paper collates fragmented findings about AI harms in insurance and organises them via TCCM to show where evidence is thin or concentrated. It argues that specific AI features act as triggers that provoke psychological reactions — chiefly concerns about fairness, anxiety about automated decisions and perceived loss of control — which then push consumers toward harmful choices (e.g. opting out, contesting outcomes, decreased uptake of beneficial products).

The TPDO (Trigger–Psychology–Decision–Outcome) framework is proposed to trace these sequential mechanisms and to identify intervention points for regulators and firms. The review also critiques the field: many studies focus on particular regions or rely heavily on certain methods, leaving gaps in theory development and empirical balance. The authors end with clear directions for future research and call for governance approaches that protect consumers as insurers scale AI use.

Context and relevance

As insurers adopt AI for pricing, underwriting and claims, this review is timely: it brings together psychological, technical and policy perspectives to show how design choices translate into real consumer harms. For regulators, insurtech teams and consumer advocates it highlights where oversight and design changes can reduce exclusion, discrimination and trust erosion. For researchers it flags gaps — the need for broader geographic coverage, mixed methods and stronger theory building — and supplies a usable framework (TPDO) for future studies.

Why should I read this?

Quick version: if you care about insurance, AI or consumer protection, this paper saves you time — it pulls together the messy literature, shows exactly how bad outcomes happen and gives a neat framework (TPDO) you can use straight away. It tells you what to worry about (opacity, over-personalisation, biased data), how customers react (angst, fairness worries, opting out) and what that means for uptake and trust. Short, sharp and useful.

Author style

Punchy: the authors don’t just list problems — they map causal pathways and make the case that AI harms are avoidable with better design and governance. If you work in regulation, product design or consumer research this is high-value reading — the TPDO framework alone is worth the time.

Source

Source: https://onlinelibrary.wiley.com/doi/10.1111/joca.70034?af=R