When More Project Transparency Can Make Team Communication Less Honest

An assistant supporting a team faces a problem that accurate message processing alone cannot solve: employees may have reasons not to reveal what they know. Providing more project context can help them make better decisions, yet also change which claims they find advantageous to make.

In their theoretical preprint, Wichinpong Park Sinchaisri and Zhiwei Wang examine this tension through a model of two workers and a firm-controlled information intermediary. Their findings concern strategic incentives in a mathematical setting, rather than measured effects from a workplace deployment.

Facts and preferences serve different purposes

Each worker privately knows a local condition and a personal preference. The local condition determines the productive direction of an action, such as increasing or decreasing activity. The preference shifts how strongly the worker wants to respond relative to the firm’s target. Workers also want their actions to coordinate with one another.

The assistant initially knows an aggregate project condition that affects the appropriate scale of those actions. In the baseline model, the firm commits to a disclosure policy before information is realized. The assistant sends public project information, workers submit simultaneous reports, and the assistant forwards those reports unchanged. Workers then choose their own actions.

Reports are costless and unverifiable. Their credibility therefore depends on whether a worker benefits from sending a different message—not on whether the assistant can faithfully reproduce it.

Why less precise context can elicit more information

The main analysis considers environments where both preference types favor the firm’s productive direction, although they disagree about response intensity. Here, workers can credibly report local conditions without revealing their preferences.

Revealing preferences requires an additional incentive condition. When the expected aggregate project state becomes sufficiently high, a worker with a more moderate target can benefit from claiming a stronger preference to influence the coworker’s action. Complete preference reporting then becomes unsustainable, even though factual reporting remains feasible.

When coordination carries relatively little weight, the firm can benefit from preserving that preference information. The authors show that pooling nearby project states into one public signal can keep reporting credible. Workers receive a less precise project forecast but supply more informative reports about themselves.

Under the stated conditions, including a uniform project-state distribution and a pooling interval that fits within its support, the authors characterize an optimal interval. The policy balances lost project precision against the additional value of preference reporting.

The result reverses when coordination is sufficiently important. Knowing a coworker’s private objective improves prediction, but also encourages accommodation of an objective the firm does not share. In the aligned setting, the firm then prefers full project disclosure paired with factual reports that leave preferences undisclosed. Information can be credible without being organizationally valuable.

Better for the firm need not mean better for workers

Workers value accommodation of their preferences more than the firm does. Under weak coordination, a firm-optimal policy can consequently improve the firm’s outcome while leaving workers worse off than direct exchange without project information.

The authors separately impose worker acceptance before private information is learned, with no compensation payments. Within a specified family of pooling policies, a binding acceptance requirement leads the firm to widen the pool. This preserves preference reporting across more project states, benefiting workers while sacrificing some of the firm’s gain. Securing acceptance and making truthful reporting strictly preferable are distinct requirements.

Timing can matter without concealing reports

Another extension releases project context after workers report. With balanced preference probabilities and the prescribed factual-reporting protocol, delayed context can sustain honest factual communication that the same information would prevent if released first. Workers still use the eventual context when choosing actions, but cannot tailor their earlier reports to its realization.

The paper also studies selective forwarding through two separate processors. Each controls only one delivery, and neither can revise both messages jointly. Under firm-preferred equilibrium selection, this arrangement can match unchanged forwarding and sometimes improve on it. That conclusion depends on the separate-decision restriction.

Numerical comparisons suggest that full initial disclosure captures most gains in the aligned environments examined. Further design matters more in some environments allowing opposing preferred directions, although the numerical searches are restricted and the compared project distributions also differ in mean and variance.

These are information-design results, not evidence of a uniquely AI capability: a human intermediary with the same information and commitment powers could implement the policies. The analysis also selects equilibria preferred by the firm; it does not establish that workers will uniquely implement them. Its central distinction is between what workers can credibly communicate and what the organization benefits from learning.