When More Employee Feedback Can Mean Less Useful Conversation

Two kinds of information, one decision

Employee evaluations help managers decide how demanding an assignment should be. But potential is not the only relevant information. Employees may also know more about customer demand, technical constraints or other project-specific circumstances.

In Feedback Design in the Shadow of Strategic Communication, Daiki Kishishita and Tomoki Tsujita investigate what happens when evaluations change how credibly employees can communicate that knowledge. Their theoretical preprint, available on arXiv, identifies a reason why more informative assessments need not produce better decisions. It does not report an experiment establishing that withholding feedback improves real workplaces.

The central mechanism is a conflict over ambition. An employee may value challenging work more than a manager does because developing skills can benefit the employee beyond the current job. A favorable assessment can strengthen that preference, widening the disagreement over which project to choose.

How the model works

The authors model three participants: a firm designing feedback, an employee with private knowledge of project circumstances, and a manager choosing the assignment’s ambition.

Employee potential has two possible levels, high and low, and initially neither employee nor manager knows which applies. The firm commits to a feedback policy that generates a signal observed by both. Full disclosure reveals potential perfectly; no disclosure leaves their beliefs unchanged; partial disclosure provides information without always resolving uncertainty.

After receiving feedback, the employee learns the project circumstances and sends a costless, unverifiable message to the manager. Economists call this “cheap talk.” The message can still be useful, but the manager must account for the employee’s incentive to push for a more ambitious assignment.

Communication therefore groups project circumstances into ranges rather than necessarily describing them precisely. More distinguishable messages allow the manager to make finer distinctions. As the employee’s expected potential rises, the preference conflict grows and fewer distinct messages may remain credible.

The researchers combine this communication model with information design: mathematically choosing feedback to minimize the firm’s expected assignment error. They compare the benefit of learning potential with the resulting change in communication about projects. Their optimization shows that an optimal policy can use at most two feedback outcomes, each associated with a particular updated belief.

Why transparency can lose

The first main result concerns full disclosure. Holding the model’s other parameters fixed, revealing potential perfectly is either optimal for every initial probability strictly between zero and one, or strictly suboptimal for every such probability. Changing how common high potential is does not, by itself, change that verdict.

Full disclosure is optimal when the employee–manager preference conflict is sufficiently small. Communication remains accurate enough that revealing potential is worthwhile. It is also optimal when conflict is sufficiently large: project communication already conveys no useful information, so feedback cannot damage it further.

Between these extremes, withholding information can improve the modeled decision. This possibility remains even when the manager and firm have identical objectives. The problem arises from the employee’s incentives, not necessarily a disagreement between management and the organization.

When full disclosure is suboptimal, no disclosure is optimal over some interval of initial beliefs. Partial disclosure strictly outperforms both alternatives for some beliefs precisely when credible communication can change its number of distinct messages across interior updated beliefs.

The optimal policy need not move steadily toward greater transparency as high potential becomes more common. The authors establish at most two separate no-disclosure intervals, allowing up to five successive policy regions alternating between partial and no disclosure.

What selective feedback reveals

In a low-prior partial-disclosure region, an optimal policy can conclusively identify low potential while leaving favorable assessments uncertain. In a high-prior region, the reverse can occur: a favorable assessment establishes high potential, while an unfavorable one remains inconclusive.

The authors interpret these patterns as relatively lenient versus stringent favorable evaluations—not rating quotas. A middle partial-disclosure region can also arise in which neither outcome establishes potential with certainty.

What the findings do not establish

These conclusions depend on a deliberately simplified environment: binary potential, uniformly distributed project circumstances, independence between those variables, and quadratic penalties for assignment errors. Feedback is public and the firm can commit to its design.

Another important assumption selects the most informative available communication equilibrium after every feedback outcome. Although that equilibrium gives all participants their highest expected payoffs among the alternatives, the model does not establish that organizations will coordinate on it.

The paper isolates a strategic communication mechanism rather than measuring workplace outcomes or incorporating the effort, learning and psychological responses discussed in other feedback research. Its lesson is conditional: evaluations can change not only what managers know, but also how much they can subsequently learn from employees.