---
id: "framework-four-channels-evidence"
type: "framework"
source_timestamps: ["§ What People Said vs. What They Really Experienced", "¶7"]
tags: ["research-methodology", "hci", "data-triangulation"]
related: ["claim-self-reports-fail", "prereq-psychophysiology"]
steps: ["\\\"Behavioral Observation: Tracking how people actually interact with the bot in real-time (e.g.", "conversation length", "pushback rates", "override attempts).\\\"", "\\\"Physiological Response: Measuring real-time bodily reactions to the system (e.g.", "skin conductance for stress/arousal", "facial electromyography for positive/negative affect).\\\"", "Work Quality Assessment: Evaluating the final output produced by the human-AI collaboration via blind expert ratings.", "\\\"Self-Reports: Collecting post-task subjective information about thoughts", "feelings", "and satisfaction directly from the user.\\\""]
sources: ["tail1"]
sourceVaultSlug: "hbr-seg-tail1"
originDay: 1
articleStem: "hbr-tail-113-ai-personality-problem"
sourceUrl: "https://hbr.org/2026/06/does-your-ai-have-a-personality-problem"
sourceTitle: "Does Your AI Have a Personality Problem?"
---
# Four Channels of HCI Evidence

To accurately capture the true impact of AI on human workers, researchers and organizations must look beyond self-reports and **triangulate data across four distinct channels**. Relying on only one channel — especially post-task surveys — creates a massive blind spot (the failure documented in [[claim-self-reports-fail]]).

1. **Behavioral Observation** — how people actually interact with the bot in real time: conversation length, pushback rates, override attempts. (This is the raw material for measuring [[concept-ai-friction|AI friction]].)
2. **Physiological Response** — real-time bodily reactions: **skin conductance** for stress/arousal, **facial electromyography (fEMG)** for positive/negative affect. Requires the background in [[prereq-psychophysiology]].
3. **Work Quality Assessment** — the final output, evaluated via **blind expert ratings** (the method behind [[claim-hostile-ai-degrades-work]]).
4. **Self-Reports** — post-task subjective thoughts, feelings, and satisfaction from the user.

The study's central methodological lesson: channels 1–3 revealed severe degradation that channel 4 alone would have entirely missed.