In my recent post on the WAR war and the electoral benefits of running more moderate candidates for political office, I wrote:

In that paper we estimated the average benefit of moderation to be about 2% of the vote (averaging over the two parties), but with the rise of political polarization it makes sense that the electoral effect of any factor would be in decline, so something more like a 1% effect seems plausible. There’s evidence that, during the past two decades, the effect of the economy on presidential elections and the incumbency advantage in congressional elections have both declined. The one factor that we’d expect to increase in importance is party balancing. Right now the Republican party controls all three branches of government (Congress, the President, and the Supreme Court), so party balancing would imply a big swing to the Democrats in any case. Whether this will be enough to overcome geography, that’s another story.

Paul Campos saw this and remarked:

Note here the anti-Schoolhouse Rock insight here that in contemporary American politics the real electoral trifecta isn’t presidency, House and Senate, but rather presidency, Congress, and the SCOTUS, which is true and more than a little sobering, given how incredibly dysfunctional elections to the Supreme Court have become.

I replied that I don’t remember Schoolhouse Rock, but I was always taught that the three branches of government were the legislative, executive, and judicial.

Bad things happen when you have only an executive function with no legislative or judicial

One thing I’ve often said is a problem with the university, and with firms in general, is that they have an executive function but no legislative or judicial.

Here’s what I wrote on the topic a few years ago:

Academic and corporate environments are characterized by an executive function with weak to zero legislative or judicial functions. That is, decisions are made based on consequences, with very few rules. Yes, we have lots of little rules and red tape, but no real rules telling the executives what to do.

Evaluating every decision based on consequences seems like it could be a good idea, but it leads to situations where wrongdoers are left in place, as in any given situation it seems like too much trouble to deal with the problem.

An analogy might be with the famous probability-matching problem. Suppose someone gives shuffles a deck with 100 cards, 70 red and 30 black, and then starts pulling out cards, one at a time, asking you to guess. You’ll maximize your expected number of correct answers by simply guessing Red, Red, Red, Red, Red, etc. In each case, that’s the right guess, but put it together and your guesses are not representative. Similarly, if for each scandal the university makes the locally optimal decision to do nothing, the result is that nothing is ever done.

This analogy is not perfect: I’m not recommending that the university sanction 30% of its profs at random–for one thing, that could be me! But it demonstrates the point that a series of individually reasonable decisions can be unreasonable in aggregate.

Anyway, one advantage of a judicial branch–or, more generally, a fact-finding institution that is separate from enforcement and policymaking–is that its members can feel free to look for truth, damn the consequences, because that’s their role.

So, instead of the university weighing the negatives of having an barely-repentant plagiarist on faculty or having the embarrassment of sanctioning a tenured professor, there can be an independent committee of the American History Association just judging the evidence.

It’s a lot easier to judge the evidence if you don’t have direct responsibility for what will be done by the evidence. Or, to put it another way, it’s easier to be a judge if you don’t also have to play the roles of jury and executioner.

Campos responded:

The answer to this supposedly is that universities now all have compliance offices of some sort when it comes to non-academic scandals, and internal mechanisms when it comes to academic ones. But if faculty governance is weak to non-existent these mechanisms are just captured by the administration. Kind of like the US government at present.

Interesting point. But I don’t think that universities (or, for that matter corporations) have had much of a legislative or judicial function even in the best of times. At most, we can rely on what John Kenneth Galbraith called “countervailing powers,” but with the disappearing of the news media and the disintegration of the government, we’re not getting much of these, either. The result is pure consequentialist reasoning on the part of the universities and corporations, with all the problems that this sort of non-rule-bound behavior can create in large organizations.

The role of the legal system

This also got me thinking about the reason for laws. OK, I’m no law professor so I’ll defer to Campos on this one, but here goes . . .

The standard characterization of the legal system, as I understand it, is that it balances on three stools:
(1) Structural ideas (fairness, stability, and various conceptions of the public good);
(2) Rule-following or, more generally, a paper trail or justification for any ruling; and
(3) Consequentialism (that is, good outcomes).

The “balls and strikes” model of judging would just use principle 2, but that won’t work, partly because rules always have some vagueness and need to be interpreted based on general principles, and partly because outcomes matter.

The “philosopher king” model of judging would just use principle 1, but that won’t work either, partly because general principles can conflict with each other and partly because different judges will interpret them differently, hence the need for some rule-following in service of some general principle of stable laws.

The “unitary executive model” (i.e., what we have at universities and corporations) just uses principle 3, but that has problems too. For one thing this creates perverse incentives (an implicit license to violate the law if it would be too costly for the government to enforce it), leading to an accretion of lawbreakers. Consequentialism also has the problem that you need to consider immediate consequences, medium-term consequences, long-term consequences, . . . there is no end. Having clear laws (principle 2) cuts that Gordian knot.

Another way of looking at this is to say that a key function of a law-based system is not just that the laws guide judges and tell people what to do, but also that the laws provide a sort of retroactive justification of consequentialist decisions.

Forward and reverse reasoning in decision analysis

This is related to the the point that decision analysis can be done in a “forward” or “reverse” fashion. In forward analysis, you state all your assumptions clearly and then evaluate the decision tree and pick the optimal decision. In reverse analysis, you apply forward analysis under a range of assumptions, and then if the recommended decision is not the decision you want to take, you go through step by step and figure out where your assumptions are wrong.

This bidirectional nature of decision-analytic reasoning is a fundamental part of classical Neumann-Morgenstern utility theory: in the forward direction, you can use utility analysis to make decisions; in the reverse direction, you can use decisions to deduce the utility function. In practice, we go back and forth: we make assumptions about probabilities and utilities in order to derive decisions, we look at patterns of decisions to infer probabilities and utilities, and when we find incoherence in our explicit or implicit patterns of decisions, we reflect on where our assumptions went wrong, and we modify our models accordingly.

The point is that coherence of decision analysis is a feature, not a bug. But coherence of decision analysis does not imply coherence of decisions. Formal decision analysis can be a method for making decisions (what I’m calling the forward application), but it can also be a method of interrogating your assumptions. In the real world we will typically do both.

Institutional decision analysis

The other thing is something that I call institutional decision analysis. The point here is that often we don’t just need to make a decision, we also need to justify the decision:

In institutional settings–for example, businesses, governments, or research organizations–decisions need to be justified, and formal decision analysis has a role to play in clarifying the relation between the assumptions required to build and apply a relevant probability model and the resulting estimates of costs and benefits.

We introduce the term “institutional decision analysis” to refer to the process of transparently setting up a probability model, utility function, and an inferential framework leading to cost estimates and decision recommendations. Depending on the institutional setting, the decision analysis can be formalized to different extents.

To get back to legal reasoning: judges can use principle 2 above (rule-following) to make decisions, and they can use principle 3 (consequences) to identify problems in the general structures of principle 1.

An important aspect of legal reasoning comes from retconning principles 1 and 2 above so as to get the desired outcomes from principle 3. And another important aspect of reasoning is to structure the goals and rules (principles 1 and 2) so that they cannot be too easily altered.

Posterior predictive checking

This is very much like Bayesian workflow, with its chaotic interplay of theorizing, model building, data collection, inference, posterior predictive checking, model expansion, etc., a pattern that, as Shalizi and I discussed, defies simplistic inductivist philosophies of inference.

Coherence as an ideal and as a point of comparison, not as a description of a world of decisions

I was also thinking about this after reading this review by Alexander Gheorghiu of a book by Ulf Hlobil and Robert Brandom. Here’s Gheorghiu:

The radical move in Reasons for Logic, Logic for Reasons is to propose a position that the authors call logical expressivism, in which logic is not about adjudicating reasoning – instead, it is a means of articulating, explaining and describing reasoning. In other words, on this account, logic is descriptive about how people reason, but not normative. The authors summarize this new position with the slogan “logic is not a canon, but an organon of reason”. . . .

The undercarriage of the book is a position Brandom has long championed: inferentialism. With roots in the work of Wittgenstein, Wilfrid Sellars, Michael Dummett and others, inferentialism is a position outlined by Brandom in Making It Explicit (1994), which argues that the meaning of a sentence is given by the role it plays in a web of justifications (rather than, say, concepts of truth).

The web of justifications; networks of conditional statements

I can’t comment on Wittgenstein and the rest; I’ll just say that I like the idea of thinking of logic as a procedure for “articulating, explaining and describing reasoning,” rather than a unidirectional process. Also, the bit about the “web of justifications” reminds me of our discussions of how a probability isn’t just a number; it’s part of a network of conditional statements; see here for further discussions and references on that point.

Again, we can see the analogy to the way that real-world legal reasoning–even ideal legal reasoning–will not follow simple forward or reverse models, those naive approaches which employ only one or two of the three principles above.