![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
On Biased Investigators
Oct. 28th, 2019 11:07 pmSo, Donald Trump and his personal defense attorney William Barr have initiated a criminal investigation into the Department of Justice's investigation of Russia's interference in the 2016 election. The theory is that all the people involved in the 2016 investigation opposed Trump's candidacy, so they made up a phony investigation to discredit his campaign and prevent him from getting elected.(*)
Thing is, there's a tiny grain of truth in this.
Let's suppose, hypothetically, that there were a government agency whose staff had dedicated their careers to objective facts, the rule of law, professional competence, and avoiding even the appearance of conflict of interest. And let's suppose that there were a political candidate who clearly had no interest whatsoever in objective facts, the rule of law, professional competence, or avoiding conflicts of interest. Predictably, most of the staff at that agency would oppose this candidate in the election. If, furthermore, they had evidence that the candidate in question was requesting and receiving substantial assistance from a hostile foreign government, in violation of U.S. law, they would be obligated to investigate the situation (and would probably become even more opposed to the candidate). And just because of the numbers, most of the people involved in this investigation would inevitably be people who personally dreaded the candidate winning the election. After the fact, the candidate could point to the preponderance of his political opponents in the investigation as "proof" that the investigation was politically biased.
There really are two a priori possible scenarios: one in which a candidate's political opponents self-select and conspire to whip up a phony investigation to discredit the candidate, and another in which a legitimate investigation is carried out by people who, for other reasons, happened to be predominantly opposed to the candidate. If one wanted to distinguish between these two scenarios objectively (as opposed to simply accusing people of the former in order to help one's political career), how would one do it?
It's a hypothesis-testing problem. I would say the null hypothesis is that people at the agency are doing their jobs impartially, while the alternative hypothesis is that there's substantial partisan bias affecting what investigations are started, how investigations proceed, and what conclusions are reported; the question is whether the observed evidence is extremely unlikely under the null hypothesis and likely under the alternative hypothesis. (Trump, if he knew about hypothesis testing, would probably start the other way: the null, default hypothesis is that everybody's conclusions are pre-determined by partisan bias, and the alternative hypothesis, to be accepted only if the null hypothesis looks ridiculous, is that people actually do their jobs.)
One objective difference is the overall political leaning of the agency: if the distribution of political beliefs in the investigation matches the distribution in the agency as a whole, the just-doing-our-jobs hypothesis is plausible, while if the distribution in the investigation is substantially more anti-Trump than that in the agency as a whole, that hypothesis becomes less plausible. But this could produce false positives, not just by chance but because those involved in the investigation might plausibly know more negative facts about the candidate than those in the rest of the agency; to avoid that, you'd need to know the political positions of people in the agency, and in the investigation, before they joined the investigation.
Another way to resolve it is to examine the evidence available at the very start of the investigation, and ask whether that level of evidence would ordinarily warrant an investigation by looking at similar cases. But that only works if there are similar cases.
Of course, there's a general rule that conspiracy theories are implausible: the bigger the conspiracy, the harder it is to make it work and keep it secret. But we're not concentrating on the (im)plausibility of the alternative hypothesis; standard procedure is to look mostly at the (im)plausibility of the null hypothesis.
What else could we use?
It occurs to me that one way to tell the difference is to see what the alleged partisans do with their information. In the case of the FBI investigation into Russia’s interference in the 2016, they didn’t go public with it until after the election, which suggests that either (a) they really weren’t doing it to change the outcome of the election, or (b) they were so mind-bogglingly incompetent that they forgot to announce it in time to have the desired effect. The former strikes me as much more plausible, but YMMV.
I'm sure the right-wing conspiracy theorists have an explanation for this too, like "they didn't actually want Hillary elected, because she would have been such a disaster for the country; they'd prefer to have Trump elected but cast doubt on the legitimacy of his election."
Edit: related question: how can you distinguish objectively between a scenario in which one party lies a lot more than another, and is correctly called out for it by non-partisan fact-checkers, and a scenario in which partisan fact-checkers accuse one party of lying a lot more than they accuse the other party?
Thing is, there's a tiny grain of truth in this.
Let's suppose, hypothetically, that there were a government agency whose staff had dedicated their careers to objective facts, the rule of law, professional competence, and avoiding even the appearance of conflict of interest. And let's suppose that there were a political candidate who clearly had no interest whatsoever in objective facts, the rule of law, professional competence, or avoiding conflicts of interest. Predictably, most of the staff at that agency would oppose this candidate in the election. If, furthermore, they had evidence that the candidate in question was requesting and receiving substantial assistance from a hostile foreign government, in violation of U.S. law, they would be obligated to investigate the situation (and would probably become even more opposed to the candidate). And just because of the numbers, most of the people involved in this investigation would inevitably be people who personally dreaded the candidate winning the election. After the fact, the candidate could point to the preponderance of his political opponents in the investigation as "proof" that the investigation was politically biased.
There really are two a priori possible scenarios: one in which a candidate's political opponents self-select and conspire to whip up a phony investigation to discredit the candidate, and another in which a legitimate investigation is carried out by people who, for other reasons, happened to be predominantly opposed to the candidate. If one wanted to distinguish between these two scenarios objectively (as opposed to simply accusing people of the former in order to help one's political career), how would one do it?
It's a hypothesis-testing problem. I would say the null hypothesis is that people at the agency are doing their jobs impartially, while the alternative hypothesis is that there's substantial partisan bias affecting what investigations are started, how investigations proceed, and what conclusions are reported; the question is whether the observed evidence is extremely unlikely under the null hypothesis and likely under the alternative hypothesis. (Trump, if he knew about hypothesis testing, would probably start the other way: the null, default hypothesis is that everybody's conclusions are pre-determined by partisan bias, and the alternative hypothesis, to be accepted only if the null hypothesis looks ridiculous, is that people actually do their jobs.)
One objective difference is the overall political leaning of the agency: if the distribution of political beliefs in the investigation matches the distribution in the agency as a whole, the just-doing-our-jobs hypothesis is plausible, while if the distribution in the investigation is substantially more anti-Trump than that in the agency as a whole, that hypothesis becomes less plausible. But this could produce false positives, not just by chance but because those involved in the investigation might plausibly know more negative facts about the candidate than those in the rest of the agency; to avoid that, you'd need to know the political positions of people in the agency, and in the investigation, before they joined the investigation.
Another way to resolve it is to examine the evidence available at the very start of the investigation, and ask whether that level of evidence would ordinarily warrant an investigation by looking at similar cases. But that only works if there are similar cases.
Of course, there's a general rule that conspiracy theories are implausible: the bigger the conspiracy, the harder it is to make it work and keep it secret. But we're not concentrating on the (im)plausibility of the alternative hypothesis; standard procedure is to look mostly at the (im)plausibility of the null hypothesis.
What else could we use?
It occurs to me that one way to tell the difference is to see what the alleged partisans do with their information. In the case of the FBI investigation into Russia’s interference in the 2016, they didn’t go public with it until after the election, which suggests that either (a) they really weren’t doing it to change the outcome of the election, or (b) they were so mind-bogglingly incompetent that they forgot to announce it in time to have the desired effect. The former strikes me as much more plausible, but YMMV.
I'm sure the right-wing conspiracy theorists have an explanation for this too, like "they didn't actually want Hillary elected, because she would have been such a disaster for the country; they'd prefer to have Trump elected but cast doubt on the legitimacy of his election."
Edit: related question: how can you distinguish objectively between a scenario in which one party lies a lot more than another, and is correctly called out for it by non-partisan fact-checkers, and a scenario in which partisan fact-checkers accuse one party of lying a lot more than they accuse the other party?
