The upside of keeping your cards close to your chest

In decision-making, ‘think twice before implementing anything like radical transparency’, research shows.

By Marina Krakovsky, based on the research of Ronen Gradwohl and Timothy Feddersen

kellog transparency
Michael Meier

You’re an executive with a big decision to make. Perhaps there is a key C-suite position to fill, or a question about whether to enter a new market. So you turn to an advisory committee, appointing people who you believe will offer wise counsel because they, collectively, have more information about the issue than you do.

It would seem logical that you would want that committee to be fully transparent with you in their deliberations, right? After all, you would get more information if you knew what went on behind closed doors.

But according to a recent analysis by a pair of Kellogg School researchers, requiring transparency may actually yield less information than allowing deliberations to go on in private.

At first blush, the intuition that transparency is best in these advisory settings makes perfect sense, explains Ronen Gradwohl, an assistant professor of managerial economics and decision sciences.

If all you receive is a committee’s conclusion, you are left with just a higher-level summary of the issue, he says. “Whereas if you have transparency, then you should get more information about the problem you’re trying to solve. And most people would think more information is better.”

But Gradwohl, who teamed up with Timothy Feddersen, a professor in the same department, points to the key flaw in this intuition: transparency means you get to see what everybody said, not necessarily everything they know.

What information you see depends on what the committee members are willing to disclose – but transparency tends to suppress the willingness to disclose information, both to the other people on the advisory committee and to the final decision-maker.

“If the committee knows that it’s going to be transparent,” Gradwohl says, “the committee members will manipulate the information they share or share less information than they would otherwise.”

You may have experienced an inkling of this phenomenon. Meeting minutes, for example, sometimes say that a particular vote was unanimous.

“Now, was it really unanimous?” Gradwohl asks, with scepticism. “It could be that people went into the meeting with differing opinions, but they wanted it to look unanimous because they know that whoever makes the decision is going to try to read into not just what the recommendation was, but also how many people were in favour, and how compelling was the evidence that this was the right decision.”

For example, a group that wishes to open a new branch of a retail chain might consist of four people who believe the new location will be very profitable and one who is unsure. After learning how certain the others are, that unsure person is likely to assume that her information was simply inferior to everyone else’s and become convinced that the new location makes sense.

But if the decision-maker knew that one person started out uncertain about the new location, he might scuttle the project.”

So to make the most compelling case to the chain’s owner, it is in the best interest of the group to claim that all five people strongly believe the new location will be lucrative.

In other words, knowing that the decision-maker will have full transparency into the recommendation process can actually change what people recommend. Moreover, knowing this, the decision-maker might be suspicious of a unanimous vote – and rightly so.

In this case, transparency fails to achieve its goal of revealing better, more accurate information.

The researchers turned to game theory to model these sorts of scenarios. Their model is predicated on the idea that the committee and decision-maker have different incentives: the decision-maker might be more conservative, for instance, because he has more “skin in the game,” while the committee may be more open to taking smart risks. Their model also assumes that the only reason decision-makers turn to a committee is because they do not have all the relevant information themselves.

In the opaque version of the game – when the only information the decision-maker is given is the committee’s final recommendation – the model shows that the committee’s decision ends up accurately representing the group’s aggregate opinion; the rational decision-maker might then find the committee’s recommendation convincing and take action on it. In this case, opacity works well, according to the researchers’ model. Any differences between the committee members and the decision-maker in terms of motivations or tolerance for risk is neutralized.

“Then we show the opposite case,” Gradwohl says, referring to the scenario of full transparency. In this case, the committee members “don’t say anything meaningful.”

This complete breakdown of communication comes about gradually, as each player second-guesses the others.

For example, consider the five committee members who want to convince the owner of the retail chain to open a new branch. Suppose that the committee members are in favour of opening a branch whenever a majority of them have convincing positive information, but that the more conservative owner is only willing to do so if all five have convincing information.

Here is what each member’s individual reasoning on the committee might look like, Gradwohl explains:

If other people lack convincing information, then what I say doesn’t matter, since we will be unable to convince the owner regardless of how convincing my information is.

The only scenario in which what I say does matter is if all four of the other committee members have convincing information. In this scenario I am also in favour of opening a branch, because their information suffices to convince me. Thus, I should tell the owner that I am in favour.

But, the owner will realize that I claim to be in favour whether or not I have convincing information, so he will essentially disregard what I say.

At the same time, other committee members are engaged in this same reasoning, Gradwohl points out.

“So at the end of the day, the owner gets five people claiming to be in favour but actually learns nothing from this because they will make that same claim regardless of their information. At the end of the day, everyone babbles.”

The committee members share so little useful information that they might as well be speaking gobbledygook.

In real life, the outcomes won’t be as drastic as in the game, Gradwohl acknowledges.
“People aren’t perfectly rational, so transparency might not lead to zero information, but it’s definitely there as a roadblock to information aggregation.”

Such scenarios – where decision-makers decide whether to demand full transparency when they seek recommendations from parties with differing incentives – are common in other contexts as well.

Consider the case of a manager debating whether to launch a risky new product. To help her decide, she turns to her employees, each of whom has information relevant to the decision. Is she likely to get better information by asking employees for their opinion one by one (analogous to transparency) or by letting the employees confer in private to come up with a joint recommendation (the opaque scenario)?

“Our result suggests that asking each employee separately would lead to less informative advice and would be inferior to letting the employees come up with a joint recommendation,” Gradwohl says.

For example, employees might be more eager than the manager to launch the new product, with less to lose if the product flops. So when the manager approaches employees individually, each employee has an incentive to exaggerate the new product’s odds of success.

But once again, the difference in incentives between the manager and the employees does not pose as much of a problem in the opaque scenario. If employees can have a private group discussion, they can freely share their information with one another before the group makes their joint recommendation.

For example, if all employees think the probability of success is low, the joint recommendation will be to not launch the new product – and the manager has less reason to doubt that the recommendation truly reflects everyone’s information. The same is true if all employees think the probability of success is high: they will recommend the launch, and the manager will accept the recommendation.

Gradwohl’s main takeaway from the research is straightforward: “Think twice before implementing anything like radical transparency.”

Contrary to conventional wisdom, backroom discussions and other private conversations can be more helpful in getting people to share information.

“These sorts of offline conversations might actually be beneficial to everybody – not just to the committee, which is obvious, but also to the eventual decision-maker,” he says.*Marina Krakovsky is the author of The Middleman Economy (Palgrave Macmillan).

This article first appeared in Kellogg Insight (