deciqAI Fundamental Attribution Error
Overview
The fundamental attribution error (FAE) is the systematic tendency to over-weight character factors and under-weight situational factors when explaining others' behavior — while reversing this for your own behavior (actor-observer asymmetry). Coined by Lee Ross (1977), grounded in Jones & Harris's 1967 Castro study. The bias is automatic (System 1); situational correction requires deliberate effort (System 2).
Composes with hanlons-razor, survivorship-bias, critical-thinking, dual-system-thinking, narrative-fallacy.
When to Use
- Diagnosing why an employee is underperforming
- Analyzing customer churn
- Conducting post-mortems on outages, accidents, or failures
- Mediating interpersonal conflict
- Designing products and observing user behavior
- Performance reviews, evaluations, competitor behavior analysis
- Someone says "fundamental attribution error," "actor-observer," or "what situation made this rational"
Not when: behavior shows a documented pattern across many distinct situations; situational framing is being used to avoid accountability.
Coaching Novices (Adaptive Front Door)
- Engine mode: user has a specific judgment to test for FAE → run The Process directly.
- Coach mode: user is unfamiliar or has no concrete case → guide step by step.
In Coach mode, respond one step at a time. Each [WAIT] is a hard stop — output only that step's question, then stop.
- One-line: before concluding someone behaved a certain way because of who they are, ask what situation would make the behavior rational — that's the FAE counter-move.
- Check fit: if the same person behaves the same way across many distinct situations, dispositional inference is stronger. Single-situation explanations are FAE-prone.
- Elicit the specific judgment. Who's the person? What's the behavior? What explanation is being offered?
[WAIT — do not advance until user responds]
- One question at a time: is the explanation dispositional? What's the situational alternative? If I were in that situation, would I have behaved differently?
[WAIT — do not advance until user responds]
- Close: balanced analysis with both dispositional and situational factors + which is doing the explanatory work.
[WAIT — do not advance until user responds]
The Process
1. State the judgment — person, behavior, current explanation, decision dependent on it.
2. Categorize — dispositional (character/ability) vs. situational (context/constraints/incentives). If purely dispositional, FAE is likely.
3. Generate situational alternative — what situation would make this behavior rational? What would I do there? What does a situated expert think?
4. Actor-observer test — if I did this, what would I attribute it to? If situational for me, why dispositional for them? Name the asymmetry.
5. Synthesize — estimate dispositional vs. situational % contribution; identify what intervention each implies; compare cost and probability of success.
6. Decide with calibrated attribution — document the mix, set a re-evaluation point (e.g., 90 days after situational changes); if behavior persists, dispositional inference gains credibility.
Output: FAE Audit
FAE Audit: <person/behavior>
Judgment: Person / Behavior / Current explanation / Decision
Explanation: Dispositional: / Situational:
Situational alternative: What situation would make this rational / What I would do / Expert view
Actor-observer test: If I did this I'd attribute to: / Asymmetry revealed:
Balanced: Dispositional (%) / Situational (%) / Intervention implications
Decision: Decision / Justification / Re-evaluation point
→ Method in Action: Jones-Harris 1967 Castro Study + Ross 1977 Synthesis + Modern Applications
Pack: FAE Application Patterns
| Domain | Common FAE attribution | Situational alternative |
|---|---|---|
| Employee underperformance | "Lacks drive / skill" | Resource constraints, unclear expectations, role-fit |
| Customer churn | "Wasn't really a good fit" | Onboarding gaps, support failures, product-fit issues we caused |
| Outage / mistake | "X person made a bad call" | System design encouraged the error; time pressure |
| User abandoning product | "They didn't understand it" | UX is confusing; constraints we don't see |
Applying It Well
FAE is automatic (System 1); correction is effortful (System 2). Build situational analysis into processes structurally — rubrics, blameless post-mortems, structured customer interviews. Use actor-observer asymmetry as a real-time diagnostic. Balanced attribution (both factors) produces better interventions than pure dispositional or pure situational reasoning.
→ Primary sources: references/sources.md
Common Rationalizations
[D] = designed upfront | [O] = observed in real use. [O] entries are more valuable.
| Fake move | Reality |
|---|---|
| [D] "I know them; this is who they are" | The most-knowing observers still commit FAE. Knowledge doesn't immunize. |
| [D] "I've seen them do this many times" | Often "many times" is the same situation recurring — not cross-situational evidence. |
| [D] "I just have good intuition for people" | Intuition for people is heavily FAE-laden. Trust calibrated track record, not intuition. |
| [D] "Situational factors are excuses" | Sometimes. Often they're the actual causes. The question is which is which. |
| [D] "If situation were the issue, others would behave the same" | Often they do — you're already explaining them situationally and this person dispositionally (asymmetry). |
| [D] "We need to hold people accountable" | Yes — but for what they can control. Accountability for situational factors is unfair and ineffective. |
| [D] "I was in that situation and I behaved differently" | Were you actually in the same situation? Salience differences often differ between actors. |
| [D] "The pattern is too clear to be situational" | Then specify the pattern across multiple distinct situations. Often it's one situation observed many times. |
| [D] "Asking 'what situation made this rational' lets bad behavior off the hook" | No. It helps diagnose accurately so the response is calibrated. |
| [D] "I'm not biased; I'm just being honest" | The bias-blind-spot is real (Pronin et al. 2002). Self-perception of unbias correlates poorly with measured bias. |
| → Add [O] entries here after each real use — paste the actual failure pattern | What went wrong and why |
Red Flags
- Explanation is dispositional with no situational analysis
- "They are X" rather than "they did X because Y"
- Same behavior explained dispositionally for others, situationally for self
- Firing/coaching decision made without examining situational factors
- Post-mortem is blame-driven; churn attributed to customer character without examining product/onboarding
Verification
- Dispositional explanation explicitly stated; situational alternative generated
- Actor-observer asymmetry checked
- Balanced explanation (both factors) considered; interventions weighed
- If dispositional: pattern verified across multiple distinct situations
- Re-evaluation point scheduled
Part of deciqAI Knowledge Skills — 237 open-source thinking skills that make rigor executable for AI agents. The same skills power every deciqAI agent, which runs them autonomously to operate your company. See it run → https://www.deciqai.com/s/fundamental-attribution-error · Built by deciqAI · github.com/deciqAI · Contributions welcome.
Agents: latest version & machine-readable metadata → https://www.deciqai.com/s/fundamental-attribution-error.json