Friday, December 2, 2016

Grappling with bias

The following was written in response to a student's paragraph on this article. The student allowed as how economists likely knew more than him, but that he didn't agree with their negative assessment of Trump's fiscal proposals, as described in the article.

He also characterized the New York Times as incredibly biased against Trump, like CBS news.

I knew that my response would be longer than it was sensible to write in my red-pen chicken scratch, so I drafted an email. It came out longer than I expected.

You're right to recognize that professional economists may have an edge on you in this discussion.

At the same time, my profession hasn't exactly covered itself in glory in terms of seeing what was coming back in 2006-08, or in analyzing it since then. Paul Krugman recognized this back around 2005, when he took questions from the audience after a talk. The question was something like, "You say that current conditions suggest a devaluation of the U.S. dollar. When do you think that will happen?" And his answer was, "According to my model, about 18 months ago." In other words, our models are imperfect. And as it's impossible to completely separate economics from politics, we economists, of various political persuasions, are vulnerable to the risk of seeing things a certain way because we want them to actually be that way.

It's a matter of balance. On the one hand, recognizing that other people may have more experience and/or expertise than you in a given area. On the other, not giving anyone a pass and accepting what they say on their authority simply because they're an expert.

The key for the layperson is to try to understand the argument a given expert is making.

The key for the expert, when talking to the general public rather than to his or her peers, is to communicate in a way that can be understood by an intelligent reader / listener who isn't already initiated into the field's way of thinking and talking.

Then all that we can hope is that the public will be open to hearing different arguments and weighing them with an open mind.

Which brings us to the subject of bias.

There's a risk that the word comes to mean, "Presenting a view I don't agree with."

But if it's going to have any useful meaning, it has to be tied to issues of whether information is true, contextualized, and proportionate.

"True" is both easy to define and hard to determine with 100% certainty. For the stuff that appears in a given day's newspaper, or a news website, there's probably absolutely nothing that I've personally witnessed. And there's probably not much in which the average reader has significant outside familiarity or expertise in order to evaluate what's been written. At one level, therefore, we are forced to somewhat rely on the honesty and integrity of the people writing the stories, to trust that they have exercised due diligence to get the facts right. If you read from multiple sources, and look for how often a given source has gotten things factually wrong in the past, and what they've done about those factual inaccuracies once they're pointed out, you can start to develop a sense for reliable information, but the process is always imperfect.

"Context" is harder. There's a tendency in the last perhaps 20 years for people to excuse really atrocious things they've said or done by saying, "That was taken out of context." Sometimes this is a valid argument. As an extreme example, suppose you had said, "If we were to reinstitute slavery, it would be better for people who owned slaves and only cared about cheap cotton, but it would be a horrific idea." If I take that statement and shorten it to, "If we were to reinstitute slavery, it would be better," I have clearly turned your meaning entirely around. I have quoted your actual words, yet I have dishonestly represented your views. Then let's say someone else comes along, sees your words in my publication and quotes them; now they're defaming you without realizing they've done it.

But "out of context" isn't an automatic "get out of jail for having said something horrid" card. It often functions that way, because people don't want to believe bad things about politicians they've decided they support, but as a matter of logic, it shouldn't work that way.

Lastly, "proportionate" is the hardest to separate from our own perceptions and preferences. Let's say I like candidate A and you like B. And let's say that you and I agree that the information presented about each is both factually accurate, and reasonably contextualized.

Presumably, both candidates have strengths and weaknesses as potential office-holders. Politics and business are not professions that tend to attract saints, so both people have probably done things that are questionable. At the same time, few people are simply bad, and so both people have probably done some good. The question then is, how much of each does a given press outlet present.

If they show exactly the same amount of bad and good about each candidate, that may seem fair, but in fact it is almost certainly a kind of bias. Consider this: what are the odds that the two candidates are actually equally competent (or incompetent), equally ethical (or unethical), equally honest (or dishonest)? Those odds are approximately zero. And so if you give "balanced" coverage, you must be putting undue weight on one candidate's negatives, or minimizing that person's positives, or putting undue weight on the other person's positives, while minimizing their negatives.

But as observers with our own preconceptions about what is real, we will judge the coverage accordingly. I will zero in on every negative about candidate A and point to it as evidence of how the media are biased against "my guy," while you do the same with every negative article about candidate B. Meanwhile, in my eyes, the negatives about candidate B aren't bias, they're just the truth, while you're singing the same song about whichever of candidate A's flaws make it into print.

So we don't get very far keeping this on the abstract level. We have to look at concrete instances.

Say there's a report that Trump advocates having U.S. personnel commit war crimes.

Presumably that's a negative characterization, and a Trump supporter might point to it as a sign of bias.

The first test is whether the statement is true. It is true that Trump advocated the use of waterboarding. And it is true that the U.S. has in the past considered waterboarding to be a war crime: after World War II we hanged some Japanese officers for involvement in a range of actions, including waterboarding.

So we can say that Trump supported waterboarding and waterboarding is arguably a war crime, and therefore, it's reasonably true to say that Trump supported having the U.S. commit war crimes.

(Actually, since the election he has apparently backed off from that position. That's why I've put the statements in the past tense - we'll have to see how his thinking on the issue evolves. But before the election, it would have been true to put those statements in the present tense.)

What about context? As described above, it's harder to pin that one down. I've looked at several links and failed to find one where a broader context suggested a different conclusion. He made numerous statements at different times during the campaign, and I didn't see a context that would have changed the meaning, whether that was additional words surrounding the advocacy, or whether he only advocated it in certain settings, which would shed some different light on his meaning. In short, I can't demonstrate appropriate contextualization as definitively as the basic truthfulness of the claim, but I don't think he was being taken out of context.

Last is proportionality, which in this case has two dimensions:
  1. Is the coverage of Trump's stance on this issue commensurate with the coverage of Clinton's stance on the same issue? and,
  2. Is the coverage of this issue in particular proportionate to coverage of other relevant issues?
That second question is fundamentally subjective. Are war crimes more important than emails, or do emails take precedence over the Trump Foundation, or is that more important than the Clinton Foundation?

The first question is easier, because the field of discussion is more contained. If a news site is going to claim that Trump supports war crimes, should it be making the same claim about Clinton?

And the answer to that is possibly "yes." Clinton supports the Obama administration's use of drones to target suspected terrorists in the Middle East. That policy has a nasty habit of killing civilians, sometimes hitting, say, a wedding, or a funeral. While the intent is to kill terrorists, the extent of what we euphemistically call "collateral damage" raises the plausible contention that the U.S. is engaged in a war crime through this policy. If you accept that judgment, then it is also true and contextualized that Clinton supports war crimes.

But then, Trump supports a similar use of drones.

So looking at proportionality on the narrower issue of whether both candidates are being treated similarly on a given question, there is arguably some bias in saying, "Trump supports U.S. commission of war crimes," and leaving it at that.

To craft a statement that would be a balanced reflection of the range of both candidate's views, you'd have to say something like, "Both candidates support an action that is possibly a war crime, while Trump also supports a further action that almost certainly is a war crime."

If we're serious about the idea of bias in our sources of information, and put in the work to try to really ferret it out, the concept can actually be a tool to help us better understand the world.

The danger of the term (and the view behind it) is that we can also use it as an easy way to avoid having to acknowledge information that conflicts with what we already believe. If we retreat to that use of the word, then we cut ourselves off from learning about the world.

No comments:

Post a Comment