The Edge of Honesty: Who Must Answer Hard Questions?
In arguments and in software development, the term “edge case” appears early and often. For those unfamiliar, the idea simply refers to the extension of an argument to the boundary of its application, beyond which the argument is unnecessary, irrelevant, or moot. This turns out to be a crucial intellectual device to refine thinking, discuss steelmanned (rather than strawmanned) positions, and reach informed conclusions. At a minimum, the exercise demands that one grapples with the implications of their position beyond the most common case for which it was conceived.
Alas, we inhabit a realm where such discourse is rare, and its selective application skews and manipulates public debate in a manner as damaging as it is invisible in many cases. For illustrative purposes, let’s dig into perhaps the most incendiary political topic for the past half-century. Oh, and remember, at no point in this conversation are we to choose and defend our position, but rather, explore and explain the boundaries of both positions.
Abortion. Let us Wade into this quagmire, wholly expecting our footwear to be sullied by the muck of controversy. We’ll wipe our feet before we re-enter polite society, don’t you worry.
But first, let’s define “steelmanning.” A common, if less-than-good-faith technique in argument is “strawmanning.” In this case, we take our opponent’s argument, restate the weakest formulation of that argument, and burn that argument to the ground (it burns easily because it’s made of straw, get it?). For instance, Jane Doe believes certain social programs lead to dependency upon the state and John Doe (no relation) then responds, “so you want poor people to suffer? You just lack compassion.” Jane’s nuanced argument is reduced to a swipe at poor people by John, who then dismisses Jane’s argument as a result. John could have steelmanned Jane by stating, “your argument suggests the need to determine unintended consequences of social programs, specifically moral hazard and weakened incentives, which might worsen the problem the policy intends to solve.” At which point, if John disagrees nonetheless, he does so having demonstrated full understanding of the argument being made.
Let us first steelman both arguments as succinctly and generously as our power of articulation and enlightened reasoning will allow:
Pro-Life: The potential in human life, be it a zygote or viable fetus is inherently precious. The destruction of that potential is tantamount to an act of murder. No human life, however developed the cellular material, is unworthy of legal protection. We do not demand that a human life, after birth, to retain sentience, intelligence, or any specific competencies to receive the protections the law provides against murder. Before birth, the same principles apply. Parental status does not permit the taking of human life.
Pro-Choice: A woman’s body must remain wholly in her control, and such, cannot be compelled to deliver a child against her wishes. As a sovereign human being of sound mind, any medical procedures she selects are her prerogative. Given the risks to a mother prenatally, the responsibilities of parenthood thereafter, and potential societal stigma throughout, it is a moral imperative to ensure a woman’s capacity to control her own reproductive future.
Now, choose the argument with which you disagree (and presumably, given the nature of the topic, you disagree vehemently). Your instinct, reasonably, might be to demand that those who hold the argument you find abhorrent defend the edge-cases of their position. Essentially, you place them upon the most challenging logical landscape their position permits, then require them to justify their position upon that unstable terrain.
Let’s try it:
Pro-Life edge case challenge: The absolute sanctity of human life requires that one protect human life even if it commences via horribly invasive acts, specifically rape and incest. “Are you prepared to tell an adolescent, raped by her father, whose emotional, educational, financial, and societal prognoses are horrific, that nonetheless, she must carry and deliver that child, even if it puts her own life at risk?”
Pro-Choice edge case challenge: Absolute autonomy permits a woman carrying a viable fetus to terminate a pregnancy as a matter of preference. “Are you prepared to allow a healthy woman, with no particular medical risks associated with pregnancy, to change her mind about maternity at the eleventh hour, require no justification, and in so doing, end the life of a voiceless human being fully able to survive outside the womb? How is a doctor, during the latter-stages of the third-trimester, reaching into a uterus to kill the nascent human inside permissible ethically?”
Given even a modicum of objectivity, one ought to find merit in both arguments and some defensible wisdom in their edge-case counterpoints. Ah, but here’s where the discourse veers into controversial territory (if it hasn’t already!)...
The pro-life edge case, in various permutations, is a common element of the mainstream discussion. Conservative congressmen are asked to reassert and clarify their positions in the cases of rape and incest publicly, frequently. And this is legitimate! Insisting that a position be supported at its edge cases is good! It allows refinement, deliberation, inspection, and this is in fact how a body of knowledge is updated with the morals of the age.
In contrast, the pro-choice edge case is almost never debated in such stark terms. How often is a liberal politician forced to wriggle in response to their edge case challenge? In this regard, the discourse is impoverished. And we, as citizens of that nation, are left to navigate philosophical questions with diminished artillery. Good debates demand the best arguments from both sides1.
If a developer failed to consider and test edge cases prior to shipping software, they’d be cast into the gorge of eternal peril2. Moreover, they would deserve their fate in a manner beyond any bridgekeeper who lacks knowledge of the airspeed velocity of unladen swallows3.
When a debate is framed without sufficient exploration of edge cases, a similar punishment is worthy, and yet, it is so rarely forthcoming. It is incumbent upon data scientists, developers, and pundits to consider the implication of their work, and how those implications extend, mathematically, morally, and financially. At AE, we steelman arguments on a daily basis, even those with which we struggle mightily in our hearts and minds. We’re better for it. Care to discuss your own positions with us?
1
Other articles (e.g. this one in the Atlantic) have asserted the need to face the opposition's best arguments, and even then, struggle to do so.
2
Considering that Python is a commonly-used programming language whose name refers to the British sketch comedy troupe of Holy Grail fame, no reason this post shouldn’t contain references to the bridge of death, the old man from Scene 24, and the ultimate price paid by those who fail to answer the 5 (er, 3) questions. Incidentally, said comedy troupe also coined the usage of “spam” to refer to unwanted mail in great supply in what is perhaps the best piece of musical comedy about canned meat ever composed.
3
African or European.