How Performance Evaluation Shapes Peer Learning Within Teams

A short summary of Hattori, K. (2026) “Performance Attribution, Knowledge Transfer, and Team Formation”.

Background

Organizations rely heavily on employees learning from one another. New-hire training, on-the-job learning, mentoring, and pair work all create opportunities for knowledge to flow from more experienced or skilled workers to their coworkers.

But this learning does not happen automatically. Teaching takes time away from one’s own work. More subtly, teaching can narrow the performance gap between the person who teaches and the person who learns.

If an evaluation system places strong weight on who individually produced how much, narrowing that gap can make teaching personally costly to the more skilled worker.

The same issue matters for team formation. Putting high- and low-skill workers together creates opportunities for knowledge transfer. But putting high-skill workers together may instead generate stronger current production because they can devote more time to their own work and benefit from complementarities with similarly skilled teammates.

This paper therefore asks two related questions: how does attributing performance to individuals rather than teams affect knowledge transfer, and how does that in turn affect who should work with whom?

What the paper does

Consider a team with one high-skill worker and one low-skill worker. Knowledge flows from the high-skill worker to the low-skill worker.

The model captures two features of knowledge transfer through communication. First, knowledge transfer uses production time for both workers. The high-skill worker spends time teaching, while the low-skill worker spends time learning. Second, transfer requires participation from both sides. The effective amount transferred is therefore limited by whichever worker wants less of it.

The workers’ outputs are complementary, so improving one worker’s productivity also raises the value of the other worker’s contribution.

The key organizational variable is the performance-attribution system. Under full team attribution, evaluation depends entirely on team output. Under full individual attribution, each worker receives credit for her own output. Intermediate systems combine the two.

Individual attribution gives the high-skill worker an attribution premium because she initially produces more than her low-skill coworker. Knowledge transfer, however, reduces exactly this output gap.

The paper then embeds this two-person team in an organization with two high-skill and two low-skill workers. The organization chooses between mixed teams, each containing one worker of each type, and assortative teams, in which high-skill workers are paired together and low-skill workers are paired together.

Finally, the paper allows the organization to choose the attribution system itself when high-skill workers have attractive outside opportunities and individualized retention pay is constrained.

What it finds

First, under full team attribution, knowledge transfer is efficient. With individual attribution, however, the high-skill worker becomes the bottleneck and transfers too little knowledge.

Under team attribution, both workers evaluate transfer through the same object: team output. They therefore agree on the efficient level. Once individual attribution is introduced, the high-skill worker also cares about preserving the output advantage that generates her attribution premium. Teaching makes the low-skill worker more productive and reduces that advantage, so the high-skill worker stops before the level that maximizes team output.

The low-skill worker would prefer more transfer, but transfer requires both workers to participate. Stronger individual attribution therefore reduces effective knowledge transfer. Beyond a threshold, transfer stops completely.

The efficient level itself does not change with the attribution rule. Attribution redistributes credit within the team; it does not change total output directly. The distortion comes from how that redistribution changes the high-skill worker’s incentive to teach.

Second, the organization switches from mixed to assortative teams before knowledge transfer would completely disappear within a mixed team.

Mixed teams create gains from cross-skill learning. Assortative teams create a different gain: with complementary production, matching workers of similar skill generates a productivity premium. Individual attribution erodes the first benefit by reducing knowledge transfer, while leaving the second intact.

As individual attribution becomes stronger, there is therefore a point at which the organization prefers assortative teams even though a fixed mixed team would still sustain positive knowledge transfer. Cross-skill learning can disappear through reorganization before it disappears through behavior within an existing team.

Team composition also creates a conflict of interests. High-skill workers always prefer assortative teams, while low-skill workers always prefer mixed teams. Output-maximizing mixed teams may therefore require organizational assignment authority rather than voluntary matching.

Third, retention pressure can itself generate stronger individual attribution.

If retention is not a concern, the organization chooses full team attribution because that maximizes current output. But when high-skill workers have attractive outside opportunities and individualized retention payments are difficult to use, individual attribution becomes a second-best retention instrument. Giving more credit to individual performance raises the high-skill worker’s payoff and can make staying more attractive.

As outside opportunities improve, the organization increases individual attribution just enough to retain high-skill workers. Knowledge transfer then falls. If retention pressure becomes sufficiently strong, the organization eventually switches to assortative teams and gives up cross-skill transfer altogether.

The same mechanism therefore creates both the benefit and the cost of individual attribution. Individual attribution helps retain high-skill workers because it rewards their output advantage, while knowledge transfer destroys part of that same advantage.

Why it matters

The first implication is that weak knowledge sharing need not come from hostility, selfishness, or a bad organizational culture. The model contains no promotion tournament, replacement threat, or future rent loss. The way current performance is attributed is enough to generate too little knowledge transfer.

A second implication is that requiring more mentoring time need not solve the problem. The model concerns effective knowledge transfer, not interaction time. Organizations can schedule meetings, require mentoring hours, or colocate workers, but they cannot easily mandate the substance of what is taught. If the underlying incentives remain unchanged, mandatory interaction may create formal contact without much additional learning while still taking time away from production.

What matters instead is compensating the worker who gives up the attribution premium created by her performance advantage. Mentoring credit, lower production targets, workload relief, promotion recognition, or training budgets can all play this role. In the model, the payment needed to restore efficient transfer is exactly tied to the attribution premium the high-skill worker gives up by teaching.

The analysis also highlights the value of compensation flexibility. If individualized retention payments are available, an organization can respond to attractive outside opportunities without distorting its evaluation system toward stronger individual attribution and sacrificing knowledge transfer.

Finally, practices that make knowledge transfer less disruptive to current production—such as documentation, standardized onboarding, pair work, and knowledge-management systems—do more than make teaching easier. They also expand the range of evaluation systems under which mixed-skill teams remain worthwhile.

Performance evaluation, knowledge transfer, retention, and team composition are therefore closely connected organizational design problems. How a firm allocates credit can affect not only how much employees teach one another, but also who works with whom and whether valuable knowledge crosses skill boundaries at all.

This research was supported by JSPS KAKENHI Grant Number 24K04912.