Human-AI Interaction & Decision Quality

Decision Acceptance Rate tracks the proportion of AI recommendations that humans act on. Automation Bias Index measures how many of those acceptances occurred without meaningful human review — the silent governance failure that most organisations have not yet instrumented. The EEOC and EU AI Act both require documented human oversight for high-risk decisions; automation bias is evidence that oversight is nominal rather than real.

• Healthcare: 79% acceptance, 31% automation bias — radiologists accepting AI diagnostic flags without independent verification (MIT CSAIL 2023)
• Financial Services: 68% acceptance, 42% automation bias — highest bias rate across sectors; credit officers approving AI-scored applications at volume without case review (BCG 2023)
• Legal: 61% acceptance, 23% bias — most conservative sector; liability exposure drives genuine review (Stanford HAI)
• Optimal zone: 65–85% acceptance with <25% automation bias. Below 50% = undertrust; above 85% = over-reliance
• DARPA XAI finding: providing explanations with AI recommendations reduces automation bias by 18% — the single most effective intervention

1. Instrument your AI systems to log whether humans accessed the explanation before accepting — this is your automation bias rate
2. Add mandatory explanation display before acceptance for P1/P0 decisions — interface friction that requires acknowledgement, not just click-through
3. Set a review SLA: for high-stakes decisions, require logged time-on-task before acceptance (>60 seconds minimum)
4. Report acceptance rates by team to leadership monthly — the act of measurement alone reduces automation bias by 12% (MIT Sloan 2022)

The automation vs augmentation split defines how AI is deployed across a decision portfolio. Automation means AI decides and acts without human involvement. Augmentation (the "centaur model") means AI advises, humans decide. The optimal split is not fixed — it varies critically by decision type, reversibility, regulatory context, and cognitive stakes. Getting this wrong in either direction destroys value: over-automation creates liability and error propagation; under-automation wastes the tool.

• Centaur model outperformance: Human+AI teams beat solo AI by 23% and solo humans by 31% on complex decisions — BCG 2023 study of 12,000 knowledge workers
• Routine decisions: 65% automation / 35% augmentation optimal — McKinsey Global Institute 2023
• Complex decisions: 25% automation / 75% augmentation — Stanford HAI recommendation
• High-stakes decisions: 8% automation / 92% augmentation — any fully automated high-stakes decision is a governance violation under NIST AI RMF
• Augmentation preference: 71% of knowledge workers prefer AI as thought-partner (McKinsey 2023); this rises to 84% for healthcare workers and 88% for legal professionals

1. Map every AI deployment to one of three tiers: Automate (routine, reversible, low-stakes), Augment (complex, consequential, regulated), Advise-only (irreversible, high-liability, safety-critical)
2. Calculate your Return on Employee (RoE): measure hours freed from automated tasks + decision quality improvement per person — this is the centaur dividend
3. Resist the automation bias in system design — the default should be augmentation, with automation requiring explicit justification and governance sign-off
4. Survey team augmentation preference quarterly — low preference scores predict adoption failure before it happens

Decision quality in human-AI systems is multi-dimensional: accuracy (correctness), speed (time-to-decision), consistency (same decision in the same context), error rate (critical failures), and cognitive load (mental effort required). The HAIS (Human-AI Integration Scale), developed by researchers at MIT and Northeastern, provides a validated 25-item instrument for measuring how well AI integration serves human cognition rather than taxing it.

• Accuracy lift: +18% average improvement in AI-assisted vs solo human decisions; up to +22% in healthcare (BCG/MIT 2023)
• Error reduction: −37% critical errors in healthcare AI with human review; −22% in financial services (Stanford HAI 2023)
• Time-to-decision: −28% faster on average; routine decisions faster by 45%; high-stakes decisions faster by only 12% (appropriate caution)
• Decision consistency: +31% improvement — AI dramatically reduces "decision fatigue" variance; humans make worse decisions in the afternoon, AI does not
• Cognitive load: −24% reduction in perceived mental effort when AI provides structured options vs open-ended assistance (HAIS scale validation studies)
• Trust Score (Edelman 2024): Global average 53/100; healthcare 62; financial services 49; legal 44

1. Deploy the HAIS scale as a quarterly 25-item survey — it takes 8 minutes and produces a validated composite score across trust, transparency, control, and explainability dimensions
2. Establish baseline accuracy and error rates before AI deployment — you cannot measure lift without a pre-AI baseline captured in the same period
3. Track decision consistency using the same case presented to the same person twice over 4 weeks — the variance is your "human inconsistency baseline" that AI should reduce
4. Monitor cognitive load as a leading indicator of adoption failure — high perceived effort predicts abandonment within 90 days, well before accuracy drops become visible