Intro to Organizational Behavior (OB): Perception and Decision Making - Common Biases, Anchoring Confirmation Overconfidence Availability Escalation of Commitment Hindsight

Common Biases in Decision-Making: Study Guide

What This Is

Cognitive biases are systematic errors in thinking that distort judgment, often unconsciously. In organizations, these biases lead to poor hiring, flawed strategy, wasted resources, and toxic cultures. For example, Netflix’s 2011 Qwikster debacle (splitting DVD and streaming services) was partly driven by overconfidence bias—executives overestimated customer loyalty and underestimated backlash, costing $40M in lost subscribers.

Key Theories & Models

• Anchoring Bias (Tversky & Kahneman, 1974): Relying too heavily on the first piece of information (the "anchor") when making decisions. Implication: In salary negotiations, the first number mentioned skews the entire discussion. Google combats this by using structured interviews with pre-set evaluation criteria to avoid anchoring on a candidate’s first impression.

• Confirmation Bias (Wason, 1960): Seeking or interpreting information that confirms preexisting beliefs while ignoring contradictory evidence. Implication: Leaders may overlook red flags in a failing project because they "know it will work." Zappos reduces this by requiring "culture fit" interviews to include diverse perspectives, not just hiring managers who "like" a candidate.

• Overconfidence Bias (Dunning-Kruger Effect): Overestimating one’s abilities, knowledge, or control over events. Implication: CEOs may pursue risky mergers (e.g., HP’s $11B Autonomy acquisition, later written off as a $8.8B loss) because they believe they can "fix" the target company. Southwest Airlines mitigates this by requiring pilots to undergo regular simulator training—even veterans—to counteract complacency.

• Availability Heuristic (Tversky & Kahneman, 1973): Judging the likelihood of events based on how easily examples come to mind. Implication: After a plane crash, companies may overinvest in safety measures while ignoring more common (but less memorable) risks like cybersecurity. Netflix uses data-driven decision-making (e.g., A/B testing) to avoid relying on anecdotal "gut feelings" about content success.

• Escalation of Commitment (Staw, 1976): Doubling down on a failing course of action to justify prior investments. Implication: Kodak’s refusal to pivot from film to digital (despite inventing the digital camera) led to bankruptcy. Amazon limits this by setting "kill criteria" for projects—if metrics aren’t met, they’re canceled (e.g., Fire Phone).

• Hindsight Bias ("I-knew-it-all-along" effect): Believing, after an event, that it was predictable. Implication: Post-mortems may unfairly blame teams for "obvious" mistakes, discouraging risk-taking. Microsoft uses "pre-mortems" (imagining failure before a project starts) to identify risks without hindsight bias.

Step-by-Step Application: How to Mitigate Biases in Organizations

1. Diagnose the Bias
2. Ask: Is this decision based on the first data point (anchoring)? Are we ignoring dissenting opinions (confirmation)? Do we think we’re "smarter" than the data (overconfidence)?

3. Example: If your team insists a new product will succeed because "it worked last time," check for availability bias (relying on a memorable success).

4. Institutionalize Checks

5. For anchoring: Use blind evaluations (e.g., removing names from resumes) or pre-commitment (e.g., setting a budget range before negotiations).
6. For confirmation bias: Assign a devil’s advocate (e.g., Intel requires teams to present counterarguments for major decisions).

7. For overconfidence: Implement pre-mortems (e.g., "Imagine this project failed—why?") or calibration training (e.g., Bridgewater Associates has employees rate their confidence in predictions, then track accuracy).

8. Leverage Data & Diversity

9. Availability bias: Replace anecdotes with data (e.g., Netflix uses algorithms to predict content success, not just executive intuition).

10. Confirmation bias: Seek diverse perspectives (e.g., Google’s "bias busters" workshops train employees to spot exclusionary language in meetings).

11. Set "Kill Criteria" for Escalation

12. Define objective exit rules before starting a project (e.g., "If user adoption is <20% after 3 months, we pivot").

13. Example: Amazon’s "two-pizza rule" (teams should be small enough to feed with two pizzas) prevents escalation by limiting sunk-cost fallacies in large groups.

14. Debias Post-Mortems

15. For hindsight bias, ask: "What information did we have at the time?" (not "What do we know now?").
16. Example: After a failed product launch, Apple reviews decisions based on pre-launch data, not post-launch outcomes.

Common Misconceptions

• Misconception: "Biases only affect irrational people." Correction: Biases are hardwired—even experts fall for them. Example: Nobel laureate Daniel Kahneman (who co-discovered many biases) admits he still makes biased decisions.

• Misconception: "More information eliminates biases." Correction: More data can worsen confirmation bias (people cherry-pick evidence). Example: Blockbuster ignored Netflix’s streaming model because they focused on DVD rental data.

• Misconception: "Escalation of commitment is just stubbornness." Correction: It’s often driven by sunk-cost fallacy (fear of wasting past investments) or social pressure (not wanting to admit failure). Example: Boeing’s 737 MAX crashes were partly due to escalation—engineers downplayed risks to avoid delays.

• Misconception: "Hindsight bias is harmless—it’s just Monday-morning quarterbacking." Correction: It distorts learning by making failures seem inevitable, discouraging future risk-taking. Example: After the 2008 financial crisis, many claimed they "saw it coming," but few acted preemptively.

Exam / Case Interview Tips

1. Spot the Bias in Scenarios
2. Common question: "A manager insists a failing project will succeed because ‘we’ve invested too much to quit.’ What bias is this?"-Escalation of commitment (sunk-cost fallacy).

3. Tricky distinction: Overconfidence (believing you’re right) vs. confirmation bias (only seeking evidence that supports your belief).

4. Link Biases to Outcomes

5. Example: "How might anchoring bias affect a salary negotiation?"-The first number mentioned (e.g., $80K) skews the entire range, even if the candidate is worth $100K.

6. Real-world case: Tesla’s 2018 "funding secured" tweet (Elon Musk’s overconfidence led to SEC fines and stock volatility).

7. Propose Debiasing Strategies

8. Exam answer structure:

   1. Name the bias (e.g., "This is availability bias—the team is overestimating risk because of a recent failure").
   2. Explain the mechanism (e.g., "They’re relying on a memorable event, not base rates").
   3. Suggest a fix (e.g., "Use historical data to assess actual failure rates").

9. Avoid the "Bias Blind Spot" Trap

10. Question: "Why might a leader who’s aware of biases still make biased decisions?"-Bias blind spot (people think they’re less biased than others). Example: Uber’s Travis Kalanick ignored warnings about toxic culture because he believed he was "data-driven."

Quick Practice Scenario

Scenario: A product team is debating whether to launch a new feature. The lead engineer says, "We should do it—our last feature was a huge success, and this one’s even better." The data scientist counters, "The data shows low user interest, but the engineer keeps citing the past success as proof." What bias is at play, and how would you address it?

Answer: Availability bias (the engineer is overweighing a memorable success). Fix: Require the team to present base-rate data (e.g., "What % of past features succeeded?") and assign a devil’s advocate to challenge assumptions.

Last-Minute Cram Sheet

1. Anchoring: First number skews decisions (e.g., salary negotiations). Not just about money—can be any first impression.
2. Confirmation bias: Seek info that confirms beliefs; ignore contradictions (e.g., hiring managers favoring candidates like themselves).
3. Overconfidence: Overestimate accuracy (e.g., Elon Musk’s Tesla production timelines). Dunning-Kruger: Least competent people are most overconfident.
4. Availability heuristic: Judge likelihood by ease of recall (e.g., overestimating plane crash risk after news coverage).
5. Escalation of commitment: Throw good money after bad (e.g., Kodak’s film obsession). Sunk-cost fallacy-stubbornness—it’s emotional attachment to past investments.
6. Hindsight bias: "I knew it all along" (e.g., blaming teams for "obvious" failures). Distorts learning—makes failures seem inevitable.
7. Debiasing tools:
8. Anchoring: Blind evaluations, pre-commitment.
9. Confirmation: Devil’s advocate, diverse teams.
10. Overconfidence: Pre-mortems, calibration training.
11. Availability: Data > anecdotes.
12. Escalation: Kill criteria, small bets.
13. Real-world examples:
14. Google: Structured interviews (anchoring).
15. Zappos: Diverse culture interviews (confirmation).
16. Southwest: Pilot training (overconfidence).
17. Netflix: A/B testing (availability).
18. Amazon: Two-pizza teams (escalation).
19. Trap: "Biases are just for bad managers."-Everyone is biased, even experts.
20. Trap: "More data fixes biases."-More data can worsen confirmation bias (people cherry-pick evidence).