Pretrained Data and Biases

In this collection, “pretrained data” is an informal way to describe the experience, knowledge and expectations we bring to a new situation. It is not a clinical diagnosis or an established psychological category.

The comparison has limits. A person has a body, relationships, changing circumstances and a lived history. Human learning continues throughout life and cannot be reduced to loading a fixed dataset into a machine. The metaphor is useful only while it helps us ask clearer questions.

For example, imagine entering a workplace where you assume that asking for help signals incompetence. You may have learned that expectation in an earlier job. In your current team, asking early might be welcomed because it prevents delays. The old expectation is understandable, but it needs checking in the new setting.

An idea learned from a parent can be accurate. An idea you developed independently can be mistaken. Familiarity, novelty and social origin do not settle a claim's truth.

You also do not need to discard a tradition to demonstrate independence. After reflection, you may keep it, change part of it or choose something else. The useful outcome is a position you can explain and reconsider when circumstances or evidence change.

People can learn by observing others as well as through direct instruction and their own experience. Psychological accounts of observational learning examine how behavior changes through observation and how this relates to a person's learning history and circumstances. Watching an action does not mean everyone will copy it or adopt the same belief about it.^[1]

When you explore a particular assumption, consider several possible routes. The examples below illustrate questions to investigate; they do not diagnose how your own beliefs developed.

Other influences are practical. The nearest shop may determine where you buy food. A shift pattern may determine when you eat. A family responsibility may limit the events you can attend. These circumstances deserve attention alongside beliefs about what is normal.

People within a family or culture can also disagree. Different generations, friendships, workplaces and personal experiences may support different expectations. “My culture believes…” often needs a more precise description of which people, practice and setting you mean.

Try writing: “In the situations I remember, this response was common.” That leaves room for other experiences without dismissing the ones you had. You can acknowledge what shaped an expectation while checking how far it applies.

A useful distinction in social-norm research separates what we believe others usually do from what we believe they approve of or expect. These are often called descriptive norms and injunctive norms. UNICEF's guidance also distinguishes such social expectations from personal attitudes and from behavior driven by other circumstances.^[2]

Several kinds of statement can appear in one conversation. Separating them helps you choose the right question.

Frequency does not establish safety, fairness or usefulness. A practice may be common and still deserve review. Equally, an unusual preference does not need to become a universal rule before you can express it.

Our impressions of a group can be mistaken. In four studies published in 1993, Prentice and Miller found that students tended to believe they were more uncomfortable with campus drinking practices than the average student was. This illustrates pluralistic ignorance: a mismatch between people's own views and what they assume others think. It was research in a particular US college setting, not a description of every community.^[3]

Before accepting “Everyone wants this,” ask whether everyone has had an opportunity to answer. Silence, participation and approval are different observations. Someone may take part because an alternative was unavailable, because they prefer to avoid an argument or because they enjoy only part of the activity.

Confirmation bias describes tendencies to favor information that supports an existing belief or expectation. It can appear in the questions we ask, the sources we choose and the standards we apply when evaluating an answer.

Imagine believing that working late is the clearest sign of commitment. You notice a colleague who stays late and completes an excellent project. You give less attention to another colleague who delivers excellent work within ordinary hours. Both cases matter to the claim, but only one easily fits your explanation.

In a classic experiment, Lord, Ross and Lepper asked 48 undergraduates with opposing views on capital punishment to assess two constructed research reports with conflicting conclusions. Participants generally evaluated the report supporting their existing position more favorably. The researchers varied which method produced which conclusion, helping reveal how the result affected judgments of the method. This illustrates biased evaluation in a specific task; it does not show that evidence always makes people more entrenched.^[4]

Rejecting weak evidence is appropriate. The concern is whether you would recognize the same weakness if the result supported your position. A balanced process can still conclude that one claim has much stronger support than another.

Nor does a defensive reaction establish confirmation bias. Someone may be tired, misunderstood, concerned about consequences or responding to a real insult. Examine the reasoning and context before assigning a psychological explanation.

Sometimes an expectation influences your actions in a way that limits opportunities to check it. Consider this invented example: a person believes that asking questions in meetings will make them look unprepared.

Several points invite a useful check. The person could ask how the team prefers to handle questions, observe what happens when someone requests clarification, or raise a manageable question in a setting where they feel comfortable doing so.

The outcome is open. The team may welcome questions, react inconsistently or actually discourage them. If the environment penalises reasonable participation, that is information about the environment. Private reflection alone cannot change an organization's incentives or behavior.

Looking at the sequence keeps the question concrete: what did I predict, what did I do, what evidence became available and what conclusion did I draw?

In everyday discussions, authority bias often refers to giving a statement too much weight because of who said it. A title, reputation or confident presentation can make a claim feel settled before its connection to the evidence has been examined.

Relevant expertise is a legitimate reason to take a source seriously. We rely on specialists because no one can independently reproduce every experiment, inspect every structure or master every field.

The National Academies' report on science communication notes that people, including scientists outside their own specialties, rely on trusted sources to interpret and validate information. It also distinguishes perceived expertise from actual credentials and discusses the roles of competence, integrity and dependability in trust.^[5]

A chef may offer excellent guidance on preparing coffee. That does not automatically make their personal caffeine routine suitable health advice for every viewer. A clinician's expertise also has a scope; a careful answer may involve checking evidence or consulting another specialty.

Financial interests deserve disclosure and scrutiny, but payment does not by itself prove that a statement is false. Look at whether the methods, findings and conclusions withstand examination. Apply the same standard to a source you admire and one you dislike.

Different roles can contribute different information. A person's experience tells you how a practice affected them. A specialist can help assess broader evidence. Your values and circumstances matter when deciding what to do with that information.

Once you learn about confirmation bias, examples in other people's arguments can become easy to spot. Your own conclusion may still feel like the obvious product of careful reasoning.

Pronin, Lin and Ross studied this asymmetry in research on the bias blind spot. Participants often regarded themselves as less susceptible to described biases than comparison people. These findings concern self-perception and particular tasks; they do not establish that every person has the same biases or that every confident judgment is mistaken.^[6]

A practical response is to make your checking process visible. “I am open-minded” is difficult to assess. “I looked for evidence that would challenge this explanation and applied the same quality checks to both results” describes something you can review.

“Be objective” is a reasonable intention, but it does not tell you what to do next. A useful checking process names the claim, gives alternatives a fair examination and makes the standard of evidence explicit.

In two experiments, Lord, Lepper and Preston found that prompting people to consider an opposing possibility had a greater corrective effect on biased judgments than a general request to be fair and unbiased. This supports a specific checking question in those tasks, not a permanent cure for bias.^[7]

Results can depend on the prompt and outcome. In a 2023 study using constructed, conflicting health-research summaries, a question about judging the same methods with opposite results reduced biased assimilation, but did not significantly change the measured before-and-after attitude shift. Generating possible flaws in a separate study did not reduce biased assimilation. A checking exercise can help with one part of judgment without changing every part.^[8]

Considering an alternative does not require giving it equal credibility. An unsupported claim does not become strong evidence because it opposes your view. Choose credible challenges and examine their substance.

For a more detailed approach to evaluating arguments, continue with Core Principles of Critical Inquiry.

A custom may carry memories, hospitality, faith, friendship or a sense of continuity. When discussing a change, name the part you want to reconsider. “I would like a lower gift budget this year” is more specific than “Our traditions are irrational.”

Begin by identifying any shared purpose. If everyone wants the gathering to feel welcoming, ask which arrangements support that aim and which create avoidable pressure. People can care about the same purpose while preferring different ways to express it.

Listen for a distinction between a factual disagreement and a difference in values. “This costs less” can be checked. “I prefer a formal meal” expresses a preference. The conversation may require practical negotiation as well as evidence.

Be equally willing to hear that a change would create difficulties you had not considered. Ask about those difficulties without treating them as proof that no change is possible.

You do not need to win an argument about the history of a custom before setting a personal limit. If a stated boundary is repeatedly ignored, describe what happened and decide what support or different arrangement you need. More explanation is not always the next useful step.

These examples are invented. Each shows a different question: whether a social expectation is shared, whether status supports a claim and whether a personal experience establishes a cause.

For the caffeine example, the FDA identifies sleep problems, anxiety and other effects as possible signs of too much caffeine, notes differences in sensitivity, and recommends gradual reduction for regular users who want to cut back. A brief “detox” is not a reliable test of a single cause: withdrawal and other changes can affect how someone feels.^[9]

Personal observations can help you ask better questions, but several factors may change together. Keep the conclusion as narrow as the observation allows. See Caffeine: The “Acceptable” Stimulant — A Deep Dive for the substance-specific discussion.

Belief revision does not always mean moving from complete agreement to complete rejection. Often, the useful change concerns scope: “This works for everyone” becomes “This helped in those circumstances, and I do not yet know how widely it applies.”

Separate three decisions: what you think is true, how confident you are and what action makes sense. You may be uncertain about why a routine is unhelpful while still deciding that its cost is too high for you.

A credible counterexample can refute an absolute statement such as “No one here welcomes questions.” It does not establish the opposite absolute statement that everyone always welcomes them. Quantities, conditions and exceptions matter.

Likewise, a majority view is not automatically correct, and a minority view is not automatically insightful. Compare the quality, relevance and independence of the evidence. A well-supported expert consensus deserves different weight from a collection of unsupported assertions.

Match effort to the decision. A low-cost preference may need little investigation. A consequential health, financial or safety decision may require relevant professional advice. You can practise independent judgment while recognizing the limits of your own knowledge.

Choose a manageable subject: a meeting habit, a spending expectation or an assumption about how someone will respond. Avoid starting with a question so broad that you cannot tell what evidence would answer it.

The prompts below are an educational exercise, not a validated psychological assessment. Write a few lines and leave unanswered questions visible.

The studies below examine particular samples, tasks and outcomes. They help explain possible patterns of judgment; they cannot diagnose the motives or reasoning of a person from a single disagreement. Examples and exercises in this article are illustrative.