Cognitive biases: six patterns to notice
A first price, a convincing speaker, a visible success story: each can shape a judgment before we examine the evidence. These six cognitive biases offer a closer look at how that happens in everyday decisions.
The Agora is an illustrated fiction with six small, interactive encounters. Walk with ordinary scrolling, or use the numbered street stops. You can leave any encounter at any time.
There is no score. One answer cannot diagnose a bias, and these six different tasks cannot be added into a meaningful mark. Each explanation names what was observed and what remains unknown.
Anchoring
This is called anchoring: an initial number, quoted before you judge, can pull the judgment that follows toward it.
Explore anchoring in the illustrated walkIn everyday decisions
Private life
The sale price
Imagine an item marked down from €200 to €120. The first price can make €120 feel like a bargain, before you compare alternatives or decide what the item is worth to you.
Corporate life
The first budget
Imagine a project first estimated at €2 million, then offered at €1.2 million. The lower figure may feel attractive because of the original estimate. Judge the investment against its requirements and expected value.
Data and AI
The first interpretation
Imagine an LLM calls a colleague's email passive-aggressive. That assessment can anchor your reading even if it is wrong. Before seeing AI's interpretation, ask: what would I think from the original evidence alone?
Exponential-growth bias
Exponential-growth bias is the tendency to underestimate growth produced by repeated multiplication, often treating it as if the same amount were added at each step.
Explore exponential-growth bias in the illustrated walkIn everyday decisions
Private life
Interest on interest
Imagine savings earning interest that stays in the account and earns interest too. The same percentage applies to an increasing balance, not just to the original deposit. Calculate the compounded balance, and check the assumptions about rates, fees and withdrawals.
Corporate life
The expanding catalogue
Imagine a team doubles a product’s independent optional features from ten to twenty. If every feature can be included or omitted and every combination is allowed, the possible configurations grow from 1,024 to 1,048,576. That is 1,024 times as many, not twice as many. Check the combinations before promising to test every configuration.
Data and AI
The branching workflow
Imagine an AI workflow with ten independent yes-or-no decisions, then twenty. If every decision sequence is possible, the number of paths grows from 1,024 to 1,048,576. A system need not execute every path on each run, but testing every possible path is a different task. Estimate what must actually be explored, rather than judging the workload from the number of decisions alone.
Survivorship bias
Survivorship bias occurs when we draw conclusions from the people, objects, or cases that remain visible while overlooking those that disappeared, failed, or never made it into the data.
Explore survivorship bias in the illustrated walkIn everyday decisions
Private life
The visible success stories
Imagine choosing a demanding fitness routine after hearing only from people who completed it. Those who stopped are missing from the success stories. Ask about their experience before judging how well the routine works.
Corporate life
The companies still standing
Imagine a team copies the practices of successful businesses while ignoring businesses that used the same practices and closed. Looking only at survivors cannot show that those practices explain success.
Data and AI
The holidays made for Instagram
Imagine an AI travel assistant learns from published travel stories. Photogenic successes remain visible, while disappointing trips that travelers never post are missing. Recommendations may favor shareable destinations over the overall experience. Ask what the unpublished experiences might reveal.
Substitution bias
Substitution bias occurs when we unconsciously replace a difficult question with an easier one and answer that instead.
Explore substitution bias in the illustrated walkIn everyday decisions
Private life
The charming first date
Imagine enjoying a confident, engaging conversation and concluding that the person is trustworthy. You may be answering “Do I enjoy talking to them?” instead of “How do they behave over time?”
Corporate life
The polished pitch
Imagine choosing an internal-system proposal because its presentation is more impressive. “Which pitch is more convincing?” is easier than “Which proposal will meet our requirements?” Check the evidence for the harder question.
Data and AI
The fluent answer
Imagine judging an AI answer by how clear and confident it sounds. Fluent writing can still be wrong. Check its claims against evidence, and keep “Is this correct?” separate from “Does this sound convincing?”
Selection bias
Selection bias occurs when the people or things we observe are not a fair picture of the whole, because of how they came to be observed.
Explore selection bias in the illustrated walkIn everyday decisions
Private life
The enthusiastic reviewers
Imagine judging a course only from reviews posted by volunteers. People with strong opinions may be more likely to respond. Their reviews need not represent everyone who took the course.
Corporate life
The optional staff survey
Imagine using an optional staff survey to describe the whole company. People with time or strong views may respond more often. Check who answered and who did not before generalizing.
Data and AI
Whose language does AI learn?
Imagine a language model trained mainly on English text. Other languages, communities and cultural perspectives may be underrepresented, making AI answers less reliable or nuanced for them. Before trusting a result, ask: what is missing from the data, and how could that affect the conclusion?
Confirmation bias
Confirmation bias favors supporting evidence over evidence that could challenge a belief. Testing only examples that fit your hypothesis may hide a broader rule.
Explore confirmation bias in the illustrated walkIn everyday decisions
Private life
The suspected slight
Imagine believing a friend is avoiding you and noticing only delayed replies. Also consider replies that came quickly and other explanations for the delays. Look for evidence that could change your interpretation.
Corporate life
The favored strategy
Imagine a team supports a new strategy and collects only encouraging customer comments. Seek objections and failed cases too, and decide in advance what evidence would make the team reconsider.
Data and AI
The favored model
Imagine showing only examples where a preferred AI model performs well. Test difficult cases and credible alternatives, and keep failures visible. A good evaluation challenges the claim.
Return to the beginning