- Description
- Un vieux projecteur de cinéma à bobines posé sur une table.
- Source
- Malcolmxl5 · original page
- Licence
- CC0 Public domain
Projection bias
We take what we feel, believe, or prefer today as the benchmark for what others—and our future selves—feel, believe, or prefer.
In brief
- What : we attribute to others (or to ourselves later) our own current tastes, emotions, motivations, and knowledge.
- Why : our present state is vivid and easy to recall; others' state—or tomorrow's—must be imagined, and we imagine it starting from our own without enough correction.
- Detect : “what motivates me motivates them,” “me, in their shoes…,” bonuses or projects designed without asking the people concerned.
- Avoid : ask rather than assume, separate your present state from the question, test with different profiles.
Definition#
Fictitious illustration
Fictitious illustration
Fictitious illustrationThe term covers two related phenomena that are best kept distinct:
- projection onto others (social projection): assuming others have the same preferences, beliefs, and emotions as we do; this is the core of the false consensus effect, described by Lee Ross and colleagues in 1977;
- temporal projection (projection bias, in the strict sense of Loewenstein, O'Donoghue, and Rabin, 2003): predicting that our future tastes will resemble our present tastes, while underestimating the effect of the current state (hunger, cold, fatigue, mood) on preferences.
In management contexts, the first is especially at play: a manager motivates, delegates, explains, and rewards the way they would want it done for themselves. The second also appears when we decide today, in the excitement of launch, what will be bearable during peak workload.
Mechanism#
- Self-anchoring. To figure out what someone else thinks, we start from what we ourselves think and adjust; the adjustment is often insufficient (see Anchoring). This is the most common explanation of the phenomenon.
- The hot-cold empathy gap. Loewenstein (2005) described how hard it is to represent an emotional or physiological state (hunger, pain, anger, fatigue) that we are not currently experiencing: under the influence of a “hot” state, we underestimate the “cold” state, and vice versa.
- Lack of information about others. We rarely see others' true motivations, but we always see our own. The selection of our relationships (we spend time with people like us) reinforces this sense of normality.
- The need for a coherent world. Thinking others are like us is reassuring and simplifies communication.
What experiments show#
In the Ross, Greene, and House (1977) study, Stanford students had to say whether they would agree to walk around campus with an “Eat at Joe's” sign for thirty minutes, then estimate how many classmates would agree. Those who agreed estimated on average that 62% of others would agree; those who refused estimated that about 33% would agree (that is, about 67% would refuse). Each side saw its own choice as the more common one.
Estimated share of students who would agree (1977)
Loading the chart…
View the data
| Agreed | Refused | |
|---|---|---|
| Had agreed | 62 | - |
| Had refused | - | 33 |
For projection over time, Read and van Leeuwen (1998) asked employees to choose a snack for the following week, either fruit or candy; those who were hungry at the time of choice more often picked the less healthy snack, even though the choice concerned their future state. Gilbert, Gill, and Wilson (2002) found a similar pattern in a supermarket: people shopping on an empty stomach bought more items not on their list than those who had eaten.
Where you encounter it#
In project management#
- Requirements gathering: the team designs the interface “the way they would like it,” for users whose job, age, or tools are entirely different.
- Documentation and training: “it's obvious” from the team's point of view, not from the user's (see The curse of knowledge, a very close bias).
- Client relationship: we assume the client wants the same thing we do (the best technical solution), when they may instead want safety, speed, or staying on budget.
- Planning: a data project manager, fresh and full of energy at the end of a cycle, plans a pace for the team that it will not sustain at the end of the semester.
- On-call and incident design: alert thresholds and rotations are set in a calm period, without putting ourselves in the state of deployment night.
In management#
- Motivation: a manager motivated by challenge and autonomy believes everyone seeks autonomy; yet one team member may want clarity and stability.
- Bonuses and recognition: we offer what we would like to receive (a dinner, a bonus, a public mention), sometimes to people who are uncomfortable being singled out.
- Delegation: “I only need an objective,” so nothing is specified; or “I need to control everything,” so everything is controlled.
- Communication: a message judged clear because we know the context, bad news announced the way we would want to hear it.
- Flexibility and schedules: we imagine others have the same family, geographic, or health constraints as we do.
- Team decisions made in a “hot” state: commitments made right after a success, or a crisis, whose consequences are measured later.
Famous cases#
“Eat at Joe's”: everyone believes others would do the same as them 1977#
Biases at play: false consensus effect, projection onto others.
What happened. Lee Ross, David Greene, and Pamela House, from Stanford University, published in 1977 in the Journal of Experimental Social Psychology a series of four studies on what they called the false consensus effect. In one of them, students were asked to carry a large advertising sign (“Eat at Joe's”) for half an hour on campus. Some agreed, others refused, and all estimated the proportion of others who would do the same. Both groups judged their own choice as more common and more “normal”; refusals were explained by personality traits of the other camp.
What this illustrates
Projection does not come from ignorance of diversity in general, but from the fact that our own response feels self-evident. A manager running the same exercise for a bonus, a schedule change, or a reorganization would also estimate that “most people” would think like them. To test, you must ask: a hypothesis about others is not data.
What motivates employees: the Kovach survey 1980-1987#
Biases at play: projection onto others in motivation, false consensus effect. Older survey, questionnaire-based, to be read as an order of magnitude.
Kenneth Kovach's survey, published in 1987 in Business Horizons, asked employees to rank ten workplace motivation factors, and asked their supervisors to rank the same factors as they imagined their employees would rank them. In the reported results, employees put interesting work and recognition for work well done at the top, while supervisors estimated that employees would put good pay and job security first. The compared ranks are very different. It is a self-reported, older survey, whose method and exact figures should be checked at the source; its value is to show a gap between what managers believe is motivating and what employees say.
What this illustrates
This gap makes it hard to motivate “from a distance.” Even if precise rankings vary by era, profession, and country, the practical rule remains valid: do not decide what motivates a team without asking them, and ask rather than reuse your own criteria as the rule.
Winter clothing ordered in cold weather 2007#
Bias at play: temporal projection (onto one's future state).
Conlin, O'Donoghue, and Vogelsang (2007) studied, using data from a U.S. mail-order seller, the link between the weather at the time of order and the return rate of winter items. Result: winter items ordered in very cold weather were returned more often than those ordered in mild weather, which is consistent with the idea that, under current cold conditions, people overestimate how much they will later use warm clothing. The authors see this as evidence of projection of the present state onto future preferences, which classic choice models do not predict.
What this illustrates
Practical translation: what we decide in the heat of the moment (after an incident, in launch excitement, on a rested Monday morning) is judged using that moment's standard. Future-committing decisions (pace, workload, client commitments) benefit from being reviewed in a cool state, for example the next day, or by asking someone who is not in the same state.
Detection signals#
- “Me, in their place, I would…” used to justify a decision that affects them.
- “Everyone wants…,” “no one likes…” with no survey or direct discussion.
- A bonus, perk, or meeting whose appeal no one has validated with the intended recipients.
- Messages are clear to the sender and trigger many questions from recipients.
- Employees with different profiles (age, profession, family situation) receive exactly the same treatment.
- Regular surprise at reactions (“I don't understand why they don't want it”).
- Major decisions made right after a strong emotional state (success, failure, fatigue) with no time to step back.
Avoidance reflexes#
- Ask rather than assume. For motivation, schedules, bonuses, workload: one-on-one interview, short survey, workshop. An open-ended question is better than a confident assumption.
- Write down your hypothesis before asking. “I think the team will prefer A.” Then compare with the responses: the gap measures your bias, and it informs future decisions.
- Test with different profiles. Before generalizing a bonus, a schedule, a tool, or training, run it by people whose situation differs from yours.
- Describe the other person using their data. Their constraints, background, goals, as they express them, rather than yours transposed onto them.
- Offer a choice. Give people several forms of recognition or flexibility, rather than one option supposedly valid for everyone.
- Decide when calm. For a high-stakes decision, wait for time to step back, or have it validated by a third party who is not in the same emotional or physical state.
- Put yourself in the future state. Before committing to a pace or workload, ask yourself: “In six weeks, at peak end-of-cycle pressure, is this still reasonable?”.
- Listen to the gap. When someone reacts differently than expected, don't conclude bad faith: ask which hypothesis about them was wrong.
With AI: detect, fix, mitigate#
AI is a good generator of different viewpoints and a good questionnaire writer, but it knows neither your teams nor your clients: its “personas” are plausible stereotypes, not data. It helps formulate the right questions, not replace real answers. And it readily reflects the assumptions you give it.
Detect#
- Identify, in a note or message, claims about “what people want” that are not based on any data.
- Flag messages whose wording assumes context the recipient does not have.
- Compare your written hypotheses with the actual responses from a survey.
Fix#
- Draft a neutral questionnaire or interview guide about motivation or preferences.
- Suggest alternative interpretations of a reaction you don't understand.
- Rewrite a message for a recipient who does not know the context.
Mitigate#
- Generate several recognition or organizational variants to submit to teams for selection.
- Have it play the recipient to test a message (with the caveat below).
- Summarize survey responses by highlighting gaps with your hypotheses.
Prompts to incorporate into your practices#
1. Question my assumptions about a team. Use before deciding on a bonus, an organizational change, or a communication.
I'm preparing [decision: bonus, schedule change, reorganization, announcement] for [team: size, roles, approximate ages, context].
My current hypothesis: [what I believe people want or fear].
1. List five reasons why this hypothesis could be wrong for some members, without using stereotypes.
2. Write ten neutral, open-ended questions to ask the team to verify.
3. Propose a simple way to collect responses anonymously.
State clearly that these are hypotheses to verify, not facts about my team.2. Test a message with a reader who has no context. Use before sending an announcement, an instruction, or documentation.
Here is a message I'm going to send: [text].
Read it as someone who has none of the following context elements: [what you know and the reader does not].
Indicate: what is not understood, undefined words or abbreviations, questions the reader will ask, and what they might misunderstand.
Propose a corrected version, then the list of assumptions my message makes about the reader.How to integrate them#
- Before a team-impacting decision: prompt 1, then send the questionnaire for real.
- Before an announcement: prompt 2, then have it reviewed by a team member who does not have the context.
- Afterward: compare your written hypotheses with the responses, and record the gap.
Precautions#
- An AI-generated “persona” is a hypothesis, not a survey: it may reproduce stereotypes (age, gender, origin). Never use it to decide the fate of a real person.
- Do not share with an unauthorized tool any personally identifiable or confidential data about your team or your clients.
- Avoid letting AI decide for you what people want: it will reinforce your own assumptions if you provide them.
- For questionnaires on sensitive topics, consult your human resources department and follow applicable consultation rules.
Limits#
- Assuming others are like us is not always an error: most people do share many preferences with us, and the false consensus effect can reflect real information (our own preferences are a clue, not a bad clue).
- False consensus and temporal projection are two distinct phenomena, often confused; their mechanisms and orders of magnitude differ.
- Lab experiments focus on simple choices (snacks, signs); applying them to team motivation is a reasonable extrapolation, not a measured result.
- Kovach's survey is old and self-reported; motivational factors vary by country, profession, and era.
- The safeguard is not to “stop assuming anything” (impossible), but to know when an assumption becomes a decision, and verify it at that point.
Sources and verification#
- Loewenstein, O'Donoghue & Rabin (2003), “Projection Bias in Predicting Future Utility”, The Quarterly Journal of Economics, 118(4)
- Ross, Greene & House (1977), “The “false consensus effect”: An egocentric bias in social perception and attribution processes”, Journal of Experimental Social Psychology
- Read & van Leeuwen (1998), “Predicting Hunger: The Effects of Appetite and Delay on Choice”, Organizational Behavior and Human Decision Processes, 76(2)
- Gilbert, Gill & Wilson (2002), “The future is now: Temporal correction in affective forecasting”, Organizational Behavior and Human Decision Processes, 88(1).
- Conlin, O'Donoghue & Vogelsang (2007), “Projection Bias in Catalog Orders”, American Economic Review, 97(4), 1217-1249
- Loewenstein (2005), “Hot-cold empathy gaps and medical decision making”, Health Psychology, 24(4, suppl.).
- Van Boven & Loewenstein (2003), “Social projection of transient drive states”, Personality and Social Psychology Bulletin, 29(9).
- Kovach (1987), “What motivates employees? Workers and supervisors give different answers”, Business Horizons, 30(5).
