In this guide
- 01Specify the decision
- 02Expose initial reasoning
- 03Seek contrary evidence
- 04Record conditions for revision
A decision bias is a recurring distortion in how information is sought, interpreted or used. In innovation work, a useful response is to improve the decision procedure: examine how estimates were formed, seek evidence that could challenge a preferred idea and record what would change the choice.
Knowing the name of a bias does not make a team impartial. Use the concepts to inspect a decision, rather than diagnose a colleague who disagrees with you. This guide complements assumption mapping and the premortem by looking at the judgement that happens before a proposal is approved.
Look for the mechanism in the decision
Tversky and Kahneman’s 1974 article on judgement under uncertainty describes availability, representativeness and adjustment from an anchor. These shortcuts can be useful, yet can also produce systematic errors. The paper does not establish that every incorrect workplace estimate is caused by one of them.
Consider a hypothetical parcel service comparing two reusable containers. A senior manager opens the discussion by saying that the larger option will probably double handling time. Nobody has yet observed a full packing cycle. Later estimates cluster around that opening suggestion.

That sequence creates a reason to investigate anchoring. It does not prove it: perhaps the manager has relevant experience that others share. Ask where the initial estimate came from, which conditions it assumes and whether people formed their own estimates before hearing it.
Keep the question attached to the work. “What evidence supports this handling-time estimate?” invites an answer. “You are anchored” invites an argument about the person. A decision memo can preserve the estimate’s assumptions so that later reviewers can see what they accepted.
Three patterns worth checking
Anchoring. An initial number or proposal becomes a reference point. In the fictional service, the first estimate may shape subsequent discussion even if it lacks a clear basis. Ask participants to explain their own reasoning before negotiating a shared estimate. An independent estimate can still be wrong; independence is not a substitute for relevant knowledge.
Availability. An easily recalled incident dominates judgement about frequency or likelihood. A vivid container failure may deserve investigation without representing the usual case. Compare the incident with an appropriate record of ordinary and difficult deliveries. Missing records should remain a limitation, not become evidence that the event is rare.
Confirmation bias. Evidence is sought or interpreted in ways that favour an existing belief. Nickerson’s review of confirmation bias examines this broad pattern across different settings. A team that interviews only enthusiastic pilot users creates a problem before it interprets any answer.
These patterns can overlap. Do not spend the whole meeting choosing the perfect label. Identify the practical vulnerability: dependence on one untested estimate, an unrepresentative incident or a search that excludes contrary observations.
Change the decision procedure
Begin by separating individual preparation from group discussion. Ask people to record an estimate, the evidence behind it and its main uncertainty before showing a preferred number. Share the reasoning as well as the result. A person who has observed the task should be able to explain why their estimate differs.

For the fictional container decision, use a short preparation sheet:
| Question | Useful response |
|---|---|
| What are we estimating? | Handling time for a defined packing and return cycle |
| What is included? | Preparation, packing, exceptions and return handling |
| What evidence exists? | Named observations or relevant operational records |
| What remains judgement? | Steps not yet tried in the proposed arrangement |
| What would change my view? | A specific observation or comparison |
| How confident should we be? | A range with reasons, rather than a precise unsupported number |
This is an original meeting aid, not a validated debiasing intervention. Its purpose is to expose the basis of judgement so the group can examine it. It does not remove uncertainty or guarantee that the final estimate is accurate.
Give contrary evidence a fair route into the room
Before collecting new evidence, write the competing explanations. A slow packing cycle might result from the container design, unfamiliarity, an incomplete kit or a difficult order. Define what each explanation would lead you to expect. That creates a more useful inquiry than collecting quotations for whichever explanation the sponsor prefers.
Use research design to match the question to the evidence. Interviews can reveal concerns and experiences; a timed task under stated conditions can examine handling work; a market-frequency estimate requires an appropriate sample. Switching between these questions during the meeting makes weak evidence look stronger than it is.
Invite someone to identify the strongest objection to the current recommendation. Give them time and access to the same material, and allow the recommendation to change. A designated challenger who is ignored after speaking contributes little. Equally, do not reward objections for their number: an objection should identify a relevant mechanism or missing observation.
Keep a record of contrary evidence and what the team did with it. If a rejected concept later succeeds elsewhere, the record should show whether the decision was unreasonable at the time or whether the circumstances changed. Hindsight alone is a poor test of decision quality.
Use structured tools without treating them as protection
A weighted decision matrix can make criteria and assumptions visible. It can also hide a preferred answer in weights, scoring rules or duplicated criteria. Examine the reasoning before interpreting the total. Test whether plausible changes in judgement alter the preferred option.
A premortem can broaden the risks a group considers, but the vividness of an imagined failure does not establish its probability. Pair it with evidence about the mechanism and a practical next inquiry. Experiment design helps turn an important concern into a question that a bounded test can address.
Morewedge and colleagues’ research on debiasing decisions studies specific training interventions. It is a reason to take procedures and training seriously, rather than assume judgement is fixed. It does not establish that reading a list of biases, using this table or appointing a challenger will eliminate bias in an innovation committee.
Evidence limits and disagreement
Research tasks differ from workplace decisions with long histories, incentives and unequal authority. A bias label may be plausible without being identifiable from a single meeting. Expertise, values, incomplete data and genuine differences in objectives can also explain disagreement.
Do not describe a colleague’s concern as irrational because it conflicts with an innovation proposal. The concern may reveal a missing operational cost. Use active listening to establish what they are claiming, then evaluate that claim against the available evidence. Facts and preferences both matter, but they play different roles.
For the next consequential choice, preserve the initial alternatives, independent reasoning, relevant contrary evidence and decision conditions. Review that record when new information arrives. The practical aim is a decision that can be examined and revised, with fewer opportunities for an unsupported first impression to become an unquestioned commitment.
Sources:
- Amos Tversky and Daniel Kahneman (1974). Judgment under Uncertainty: Heuristics and Biases.
- Raymond S. Nickerson (1998). Confirmation Bias: A Ubiquitous Phenomenon in Many Guises.
- Carey K. Morewedge and colleagues (2015). Debiasing Decisions: Improved Decision Making With a Single Training Intervention.



