Egocentric Bias: Why Your Share of the Work Feels Bigger

You witness all of your own work and a sample of everyone else's. Every hour you spend on a shared task is encoded from the inside, with the effort, the interruptions and the decisions attached; the hours your colleagues put in are visible only as results, and only when you happen to be looking. When the question comes later, memory supplies what it has. Your contributions come back quickly and in detail, the other person's come back slowly and in outline, and the ratio between them feels like a measurement.

Michael Ross and Fiore Sicoly called this an availability account. The bias does not require anyone to think well of themselves, and it is not only about credit: in their data, people also took a disproportionate share of responsibility for things that had gone badly, which a straightforward self-flattery account does not predict. What the memory does is overweight whatever it holds most of, and what it holds most of is you.

The arithmetic makes the error easy to see from outside and impossible to see from inside. If four people each honestly report the share they remember doing, the four reports can add to 130 or 140 percent without a single person exaggerating on purpose. Each one is reporting a real memory. Nobody in the room has the other three memories, so nobody can tell that the total is impossible until someone writes the four numbers down together.

On this account the size of the error should track how much of the other person's work you can see, because work done out of your sight cannot be remembered and counted while every hour of your own can. What has been tested directly is prompting: when people are asked to list what each other member contributed before estimating their own share, the estimates come down, which is the most useful fact in this literature and the basis of the counter-measures below.

Michael Ross and Fiore Sicoly published "Egocentric biases in availability and attribution" in the Journal of Personality and Social Psychology in 1979. In the first study, both members of 37 married couples were asked separately to mark how much of each of 20 household and relationship activities they were responsible for: cleaning the house, making breakfast, washing dishes, planning joint activities, starting arguments, resolving them. The mark went on a 150-millimeter line whose two ends were labeled primarily wife and primarily husband, and neither partner saw the other's answers.

Adding a couple's two marks together gives a figure that should come to one whole line for an activity they divide, whatever the split. For 16 of the 20 activities the two marks summed to more than the whole, and 27 of the 37 couples overestimated across the set. The excess appeared for the unwelcome items as well as the welcome ones, which is why the authors described the result as a memory effect rather than a self-flattery effect. The paper notes that percentage estimates had been used while piloting the study and were dropped, because comparing the two percentages afterwards set the spouses arguing. They also found that when participants listed specific examples of contributions, they produced more of their own, 10.9 on average, than of their partner's, 8.1, which is the availability mechanism showing up directly.

Later studies in the same paper carried the pattern out of the home: members of basketball teams recalled their own part in the important turning points of recent games, and participants in laboratory discussion groups claimed a larger share of the group's contribution than their teammates granted them.

The unpacking repair has been tested since. Kenneth Savitsky, Leaf Van Boven, Nicholas Epley and Wayne Wight reported in 2005 that asking group members to consider each other member's contribution before giving their own figure brought the totals down toward 100. Eugene Caruso, Nicholas Epley and Max Bazerman reported the same pattern in work groups in 2006 and noted the cost of the repair: the members who had contributed most became less satisfied with the group and less interested in working with it again than the members who had contributed little, and most of all in cooperative groups.

Any effort that is shared and then discussed afterward is a candidate.

The bias lives in what memory makes available, so the repair is to make the other contributions available before the question is asked.

• Unpack the other side first. Before you say what share is yours, write down what each other person did, in the same detail you would use for yourself. This is the manipulation that reduced the totals in the 2005 and 2006 studies.
• Keep a record while the work happens. A shared log beats four good memories, because the log samples everyone's hours and memory samples only your own.
• Collect the numbers together. Ask everyone for a percentage, add them up, and show the total. The sum does the arguing, and nobody has to be accused of anything.
• Expect the pattern in blame as well as credit. When something goes wrong, the person closest to it tends to claim more of it than the record supports; map who was involved at each decision before anyone volunteers.
• Be ready for the cost. Seeing the others' contributions clearly can lower how good the group feels, which is a reason to do it early and in writing rather than in the middle of a dispute.
• Remember the limit of the arithmetic. Someone really can be doing 60 percent. The totals only show that, on average, the claims cannot all be right.

Several biases put the self at the center of a judgment. They differ in what is being overweighted.

• It is not the self-serving bias, which takes credit for success and pushes blame for failure outward. Egocentric bias runs in both directions, and Ross and Sicoly found inflated self-attribution for unwelcome activities as well as welcome ones.
• It is not the spotlight effect, which is overestimating how much other people notice you. That is a judgment about their attention; this is a judgment about your own contribution.
• It is not the false consensus effect, which is overestimating how many people share your view. Both start from the self as the default reference point, but false consensus projects your position outward rather than enlarging your share.