Tiebreaker Reasons

The beginning of the term affords opportunities to think about things I normally don’t think about, so here is a topic brought by reading discussions of hiring practices: tiebreaker reasons.

What are tiebreaker reasons? They are the reasons that determines an agent’s decision or judgment when all other reasons are equal. For the intuitive notion, consider the following example. When one says, “Our final two candidates, First and Second, are as good as each other with respect to their research, teaching, and service, but we should hire First because she is from Winnipeg,” one is offering being from Winnipeg as a tiebreaker reason for hiring First over Second. As the name indicates, intuitively tiebreaker reasons should only matter when there is a tie.

Tiebreaker reasons like that one are, I think, often offered in casual conversations. But I worry: are there really tiebreaker reasons? how should tiebreaker reasons be modeled? and ultimately, are tiebreaker reasons epistemically rational to have?

To begin, consider an initially plausible, but ultimately unsuccessful, way of modeling tiebreaker reasons. Let’s say we give all reasons some epistemic weight with respect to a decision or judgment–roughly, this measurement includes factors such as how epistemically good a reason is and how relevant it is. One might think to model tiebreaker reasons as reasons that are given infinitesimally small epistemic weight ε. This way, I think, ends up running against the intuitive notion of tiebreaker reasons because it sometimes makes tiebreaker reasons matter even when there is no tie. Consider a modification of the hiring case above. Suppose research, teaching, and service considerations give us reasons weighted N in favor of Second and reasons weighted N-ε in favor of First. Then it seems that though there’s previously no tie, giving epistemic weight to the fact that First is from Winnipeg results in a tie. It appears that being from Winnipeg no longer deserves to be called a tiebreaker reason. So this infinitesimally small way of modeling tiebreaker reasons cannot be quite right.

Its failure suggests a different way of modeling tiebreaker reasons that better fits our intuitive notion. It is not enough to give tiebreaker reasons infinitesimally small weight, it must be given no weight when there is no tie. That is, we could put in a ceteris paribus clause: all else being equal, assign the reason some arbitrary non-zero epistemic weight; else, assign it zero epistemic weight. Although this ceteris paribus way of modeling tiebreaker reasons seems successful in capturing the intuitive notion, I now worry about whether it makes tiebreaker reasons irrational, in the thick epistemic sense, for an agent to have.

Having a tiebreaker reason seems irrational because it means that the agent has to assign that reason an epistemic weight independent of factors such as how epistemically good that reason is and how relevant it is. Put this concern in the form of a question: why should what other reasons an agent has, if they have no direct bearings, matter for what epistemic weight she assign to a reason? That seems arbitrary and unjustified.

Perhaps there is another way to model tiebreaker reasons that both preserves the intuitive notion and does not make them irrational to have, but I have just overlooked it. Or perhaps, just as the ideal text of physical laws does not include any ceteris paribus laws*, the ideally rational agent’s list of reasons does not include any tiebreaker reasons. There are, literally speaking, no tiebreaker reasons; what we put forth in casual conversation as tiebreaker reasons are best understood as hinting at underlying non-tiebreaking reasons–perhaps reasons of the infinitesimally small weight sort.

(I do not mean to say the strong conclusion is supported by what I say earlier; just throwing an idea out. The post is probably best read as a train of thought on something I find interesting rather than an argument. Discuss! As always, comments that clue me in to relevant discussions would be welcomed!)

* Cheeky point thrown in to make the post more controversial.

Advertisements

6 Responses to Tiebreaker Reasons

  1. […] note on tie-breaking reasons January 15, 2009 Interesting post from Sam over at Go Grue. In reply, a quick, wholly unoriginal decision-theoretic argument that […]

  2. Steve C. says:

    Interesting post, Sam. The way you’ve set things up (“tiebreaker reasons should only matter when there is a tie”), it seems like something closer to the ceteris paribus interp is the way to go: a fact of this sort is no reason at all unless there is a tie. (I’m assuming that there is no such thing as a weightless reason. Controversial?)

    Perhaps it is best to think of this in terms of the lexical priority of one set of considerations (features of candidates relevant to job performance) over another set of considerations (desirable features irrelevant to job performance). If you beat me out at the first-level (even if by a very narrow margin), no amount of second-level success on my part makes it the case that I ought to be hired over you.

  3. Steve C. says:

    By the way, I’m not seeing why having tie-breaking reasons in this sense is irrational. Suppose that a hiring department (i) must hire someone who can teach course 100, and (ii) would really love to hire someone who can teach any of the courses ranging from 101-9,999. Strangely enough, they have only 3 applicants: Jon, Jen, and Jan.

    Jon can teach 100 only.
    Jen can teach 100 and 101 only.
    Jan can teach 101-9,999 only.

    Jen gets hired. In the minds of the hiring committee members, the fact that Jen can teach 101 serves as a reason to hire her over Jon, even though the fact that Jan can teach 101 (along with 102-9,999) does not count a bit in favor of hiring her. Is the suggestion that it’s irrational for the dept to treat things this way? (I don’t quite follow the paragraph where you express the irrationality worry, so perhaps I’m misunderstanding your point.)

  4. Shen-yi Liao says:

    Steve,

    In your case, I am not seeing what is the tiebreaking reason. As I see it here are the reasons and their weights for hiring a candidate: teach 100 – weigh that 1; teach 101 — weigh that .1. So there’s 1.1 weighted reasons to hire Jen, 1 to hire Jon, .1 to hire Jan. (Obviously all this talk of weights is oversimplistic, but bear with me.) I think the case can be understood such that the candidate can teach 101 is a lesser reason, but not a tiebreaking reason.

    But I think you are right (and Nate’s post brings this out too) that one possibility I have not considered is to model tiebreaking reasons as second-order rules for reasoning. At least, it’s not clear to me whether the ceteris paribus interpretation is something along those lines.

    My irrationality complaint is about the arbitrariness of assigning some reasons weight independent of typical factors. I think the move from arbitrary to irrational is weak. Can you think of a way to push the point a little stronger?

  5. Steve C. says:

    Well, I don’t think the “lesser reason” approach you suggest can work. You assign .1 weight to the ability to teach 101. But since the ability to teach any of 102-9,999 are just as highly valued, ability to teach any of these should also be given .1 weight. Thus, on that picture, Jan’s “score” is actually 989.9. (Perhaps it’s less for reasons of diminishing marginal utility, but I think it’d still be a landslide in Jan’s favor.) But clearly Jan cannot be the one they should hire since they must hire someone who can teach 100. If the lesser-reason picture implies that she ought to be hired, that picture can’t be right.

    Yes, I think you’re right that Jen’s ability to teach 101 isn’t a good “tie-breaking” example (since that seems pretty relevant to job performance). We can alter the case and make the second-tier considerations a collection of properties that are irrelevant to job performance (where they’re from, how nice they smell, their juggling ability, etc.).

  6. Shen-yi Liao says:

    I chose the values arbitrarily just for illustration, but you’re right that .1 is inappropriate. Choose ε instead. At any rate, even if we are giving only infinitesimally small epistemic weight to the reasons, that is intuitively enough to make them not tiebreaking reasons.

    More generally, I am not quite worried about what the reasons are, but what weights we’re assigning to them. In this regard, I think it may be helpful to step away from real hiring cases. It may be the case, especially if I am right, that no one making hiring decisions are really offering tiebreaking reasons.

%d bloggers like this: