How biased advice can be good advice

Confirmation hearings for Donald Trump’s cabinet nominees are getting underway, and many commentators have expressed concern over the lack of ideological and political diversity among Trump’s cabinet nominees and advisors. I don’t want to get in to the specifics of these individuals’ qualifications, but I do want to examine an underlying assumption behind worries about ideological homogeneity in the incoming administration.

There is an idea, summarized by the expression “team of rivals,” that great leaders surround themselves with advisors with very different views of the world. Great leaders, according to this idea, incorporate a diversity of perspectives among their advisors to counteract their own biases rather than surround themselves with an echo chamber of like-minded individuals.

The team of rivals idea is only half-true, however. It’s true that good advisors should not falsify or misrepresent their beliefs to leaders. Good advisors need to have integrity and honesty, and flatterers generally make poor advisors as Machiavelli emphasized.

However, good advisors should not necessarily be unbiased or have very different biases* from the leader they advise. It is often optimal for leaders to solicit advisors with biases similar to the leader’s, or simply advisors with a bias rather than unbiased ones. The advantage of biased advisors is even stronger when there is urgency or information is costly (as all interesting decisions are). The optimal level of bias is not zero–instead, the optimal level of bias is that level most likely to shift the leader’s decision from what the leader would choose in the absence of advice.

In his 1985 paper “The Value of Biased Information: A Rational Choice Model of Political Advice,” Randall Calvert explains this counter-intuitive claim. The intuition here is that advice from someone who has a very different bias from you is a less credible source of information and less likely to change your own views. It is hard to tell with such an advisor how much of the recommendation reflects the true probability of the action’s success, and how much it reflects the advisor’s own bias. However, if someone with a similar bias to you recommends against the action toward which you’re both biased, or in favor of the action you’re biased against, that sends a very credible message.

An advisor with the same biases as a leader is more likely to advise actions consistent with the leader’s bias, which is a cost. But when the advisor recommends against his bias, that recommendation is much more credible. Put differently, suppose the advice in question concerns whether to stick with the status quo or adopt an alternative. Suppose the advisor and the leader are both biased in favor of the status quo. If the advisor recommends accepting the alternative, this is far more valuable to the leader a recommendation to accept the alternative from a neutral advisor or an advisor biased in favor of the alternative. The value of this information–a recommendation against one’s own bias–outweighs the cost of the advisor’s own status quo bias. Such a recommendation may eliminate the need for the leader to seek out further advisors and the cost of gathering further information.

Of course, there are also other dangers to advisors who share a leader’s own biases as explained above. But as Calvert showed, the optimal level of bias in an advisor is not zero for rational decision-makers. We are often better off receiving advice from people who share our own biases when information is costly and time is scarce.

*I should note that “bias” here refers to different things in the case of the leader and the advisor. In the leader’s case, a bias represents a prior belief that a particular action will be good or bad. In the advisor’s case, the advisor makes recommendations about an action that’s imperfectly based on the true probability of the action’s success. The advisor’s bias is how much the advisor discounts the true probability of success before recommending the action.

 

 

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s