A Defense of Offense-Defense: Replying to Bogaty and Gosain

By Akhil Jalan (with significant input from Arjun Tambe and David Dosch)

Los Angeles Debate Intensive

Bogaty and Gosain criticize the offense-defense paradigm here: http://nsdupdate.com/2016/03/16/an-alternative-view-of-offensedefense-by-michael-bogaty/

In this article, I will outline what I mean by an offense-defense paradigm and answer Bogaty and Gosain’s criticisms of the view. My main criticism is that there is indeed, as Bogaty and Gosain anticipate, no fair way to determine what counts as terminal defense and what is mitigatory.

In order of importance, I think one should read sections IV, III, II, and I. IV and III are where I directly answer Bogaty and Gosain, while I and II deal with broader questions of theory paradigms and the goal of these paradigms to begin with.

Part I: What are Our Views?

Bogaty and Gosain, to my understanding, defend a modified-version of offense-defense that counts more defensive arguments as terminal. They write, “To win an argument, debaters must answer and defeat responses that indict the validity of their link chain. This does not eliminate the distinction between mitigatory and terminal defense; it merely changes it to something more logical. Any argument that indicts a fundamental premise of an argument is terminal defense and, if won, reduces one’s credence in the argument to 0%. An argument that lessens the severity of an impact is mitigatory defense and plays out exactly as offense defense does now.” [1]

The ambiguity of what Bogaty and Gosain might mean by “fundamental premise of an argument” is, of course, the key issue at hand. I will outline the problem later in this section.

Instead, I believe we should instead use a more traditional conception of offense-defense, in which defensive responses lower the judge’s credence in a particular premise (and by extension, the credence of the entire argument). For the rest of this section, I will explain what I mean by offense-defense as a paradigm for debate.

Roughly, what I mean by offense defense is the view that the judge should vote for whatever side they have most reason to vote for. [2] There are 2 kinds of things which offense-defense is concerned with: probability of an outcome and value of that outcome. [3]

This two-fold concern is best illustrated by classical rational decision theory – specifically, the VNM utility function [4][5]. While it sounds like a mouthful, this simple paradigm is one that most people employ without even realizing it. In a nutshell, it says: take the probabilities of each outcome of an event multiplied by the value of that outcome, and add it up.

Say I am offered a drink from an enemy of mine, which I am 1% sure contains poison and 99% sure is harmless. Would I drink it? VNM says no, since the harm from dying from poison is more than 100 times worse than the benefit from having a tasty drink.

In economics and decision theory, VNM is a description of what an idealized agent would do. Of course, in real life people don’t have a complete list of probabilities and often act using mental shortcuts and biases. But if a person knew what the likelihood of each result of some action was, and they assigned values to each of those outcomes (for example – tasty drink, +1 units, death,  -1,000 units), it seems they could then make the optimal decision.

In debate, an offense-defense judge should approximate this model of decision-making by considering what values they have (perhaps they have a 100% confidence that one should treat people as ends in themselves as a result of the framework debate) and what relevant values each outcome of their decision reflects: usually, the judge compares the outcomes of the aff world and the neg world. Offense-defense means the judge should adjust their credence level in a particular argument (their belief in its likelihood of being true) on a 0-100% scale based on arguments made in round.

Returning to the problem of what counts as “logical flaw,” this new probability model of argument illustrates the problem with Bogaty and Gosain’s view. Terminal defense, in this model, would be an argument that concludes there is a 0% probability of an argument being true. Equivalently, we would be certain that the argument is false. But how can we be certain of anything in the real-world, when most debate arguments are based on empirical observation and extrapolation from data? Even philosophical arguments, which sometimes attempt to be pure extensions of logic, are arguably based on base assumptions such as “humans deserve autonomy” which we can very fairly claim we don’t know for certain [6]. Compounding the problem, it’s unclear how judges can make this judgment call, since every defensive response that correctly attacks a premise in an argument can be said to point out a “flaw in the logic” of the argument as Bogaty and Gosain say.

In particular, this means that a rational-decision maker (and by extension, a rational judge) should pick the side with the highest net value, even if that side’s arguments have a lower probability (recall the poison example). This seems to be my main disagreement with Bogaty and Gosain, who argue that certain defensive responses are under-valued by an offense-defense paradigm. But the question of what defense should be relevant carries assumptions about what the nature of debate should be, leading to the next section.

Part II: Rational Decision-Making makes Debate Better

In this section I argue that debate should be an open game whose evaluation is centered only on rational decision-making. Quite simply, this means we should not attempt to encourage or discourage the development of certain argument styles (fast, slow, kritik, philosophy, etc) but rather set certain rules for argument evaluation and then let things play out on their own.

How should we facilitate this? Judges, I argue, should strive to vote for the argument that is the most true [7]. Bogaty and Gosain might argue that they believe the same, but the common thread of their arguments seems to appeal to some undefined notion of “usefulness” or “real-world-ness” that I believe would support their arguments even in light of these criticisms. They specifically mention the example of “a defender of the policy in the real world” [1] and claim their view would encourage “substantive arguments near the center of the topic literature” [1]. In the last section, I have outlined why it is clear that the offense-defense paradigm reflects the deliberative process that an actual decision-maker ought to make if they want to pick the action they have most reason to pick, a reflection of VNM utility. Bogaty and Gosain’s view would, in effect, discount arguments below a threshold of probability even when it is irrational to do so.

In effect, their view would result in encouraging judges to vote for the side of less truth.

A.   Rational Decision-Making Outweighs Real-World Skills

One might object that debaters go on to use their public speaking and advocacy skills in the real world. In such situations, debaters need to be able to communicate their arguments to other people effectively and convince them that they’re right. Offense-defense and rational choice theory are too complicated and technical to explain in a courtroom or boardroom argument; understanding how actual people think is much more useful.

But this seems anathema to the whole project of flow-based national circuit debate, which is to find out who really won the argument as opposed to who sounded better. What defines an “intelligent response,” according to Bogaty and Gosain, is simply a collection of intuitions about what is and isn’t persuasive to the authors themselves. They decry “absurd impact turns” at the expense of “smart analytics.” Further, they argue that in the real world, someone would respond to the accusation that a policy is racist by claiming, “no, it’s not.” Surely this is not up to the standards of a complete argument in current LD – and strangely, it seems Bogaty and Gosain prefer that brushing aside arguments be a sufficient form of defense. I believe if such arguments are so silly then debaters should be able to find an easy way to outweigh them with other sources of offense; the addition of paradigmatic censorship is unnecessary.

Further, in response to the oft-repeated claim that the so-called esoteric arguments in debate are not “real-world,” I would argue that the skill of being able to find and argue about such arguments is a crucial educational value of debate. Complex and nuanced evaluation of issues is an essential ability for debaters entering the “real world,” having to explain issues such as business, politics, and more in the context of day-to-day interactions. In these domains the skill of looking for implausible and unintuitive outcomes is an asset, not a detriment; debaters more able to pay attention to the details that Bogaty and Gosain dismiss as too improbable will be far more likely to succeed.

B.   What Counts as Clash

If offense-defense encourages “stupid arguments,” Bogaty and Gosain’s solution would be to grant debaters an easy way of dismissing them instead of debating them normally, in effect encouraging their total elimination from debate. While it is an open question whether clash on these kinds of arguments would be valuable, it is clear that Bogaty and Gosain’s view would limit, not expand, the types of arguments that are accessible to debaters.

In a big picture sense, this is clear because Bogaty and Gosain seem to accept only part of the rules of traditional offense-defense, but heavily advantage defense against improbable arguments. This would likely herald a return to simple, stock arguments whose probability is fairly certain. I would argue this year’s handgun ban topic is an excellent illustration of why this style of debate would be awful: arguments are stale, repetitive, and the limited topic domain leaves little room for innovation. In modern LD, where far too many debates are won on recycled theory, frameworks, and Kritiks, we should take every opportunity to expand argumentative possibilities even if initially implausible.

Further, what does it say about the state of our activity that there is a supposed wealth of “stupid arguments” which somehow keep winning? Are our judges being continually hoodwinked? I think, rather, that Bogaty and Gosain are throwing out the baby with the bathwater; their case studies are simply instances of other issues in LD that should not result in a wholesale modification of the offense-defense paradigm itself.

Part III: Case Studies

AI Disad

I would first point out that none of the 3 objections that Bogaty and Gosain outline here are an attack on the logical structure of an argument. “Logic” is a term that I believe is misused here, since logic concerns statements that can only be either true or false and whose certainty we can verify with reason entirely. In the real world, issues of technological development and the likelihood of war are firmly in the realm of probability; however unlikely, there is no logical reason that the DA in question would be false, since the future is always uncertain.

In their footnote here, Bogaty and Gosain write, “The distinction should be clear. Terminal defense points out a flaw in the logic of an argument while mitigatory defense accepts the argument as a logical possibility and attacks its probability or magnitude.” Unfortunately it is far from clear to me what a flaw “in the logic of the argument” means. Each argument has far too many premises to be stated in a finite amount of speech time, so pointing out missing links, while mitigatory, does not qualify as terminal defense here.

I would suggest that in situations like these debaters simply point out that when working at extremely low probabilities (in the event that a killer AI develops, for example) we cannot know whether outcomes are positive or negative. If we are operating at the level of extremely low probabilities, there is likely an equally small chance that an AI would be a benevolent overlord as the likelihood of AI apocalypse.

Funding Spec

This scenario is mainly an issue of reasonability and competing interpretations in theory debate. Reasonability is clearly the consequence of Bogaty and Gosain’s view of offense-defense to theory, since it lowers the bar for what counts as terminal defense on theory. I personally happen to agree that this view of reasonability is compatible with competing interpretations (indeed, I think it is the only view of reasonability that makes any sense), but it is separate from the issue of this article as a whole.

Issues of reasonability and competing interpretations are worth discussion, but that deserves its own article. For now, I will simply say that one can accept reasonability in by lowering the bar for terminal defense in the context of theory, while also accepting my view of offense-defense elsewhere on substance.

K No Links

I believe this stems from an up-layering understanding of Kritiks as opposed to viewing them as arguments that operate at the same layer as the affirmative case itself. The latter view resolves this issue because any miniscule offense from the case would outweigh the marginal link to the K (assuming the affirmative has made a permutation).

Part IV: Responses to Common Objections

A: These Low Probability Arguments are Impossible to Beat Otherwise

First, outweighing such low probability arguments should be easy if one has offense on the same layer. This was pointed out in the Kritik case study above. In another example, consider a low-probability extinction level DA. One can easily argue that if, for example, the neg claims the aff leads to economic collapse and nuclear war, that one can claim that a massive economic downturn might lead to war but a short economic downturn would be good since the government would be more reluctant to expend resources on a war. In general, it’s easy to think of such turns against poorly constructed Disads.

Second, at low probabilities one can simply argue we don’t know what direction the links will go in. The likelihood of massive oil shocks, for example, is so low that we have no idea what would happen afterwards, and even the neg’s so-called experts are simply making guesses. This also deals with the often-claimed issue of not having enough evidence.

B: How Do We Deal with Cheap-Shot Theory?

As mentioned in the reasonability case study above, this is an issue outside of the scope of this article. Reasonability is compatible with my view of offense-defense as well, since I think it is simply a way of evaluating theory that we would not use on substance. There are great arguments about why one might want to do this, with the proliferation of cheap-shot theory being among the best.

Part V: Conclusion

In conclusion, there is no non-arbitrary way to distinguish between terminal and mitigatory defense. I argue that most of Bogaty and Gosain’s case examples point to other systemic flaws in LD or are borne out of a negative gut reaction to certain argument styles rather than a careful consideration of the arguments being made. While it is certainly fair to believe some arguments are better than others, the creation of theory paradigms that facilitate this is dangerous. We should be very hesitant to essentially censor argument types by designing the rules of the game to exclude them from the beginning. This discussion of offense-defense is thus crucial to the direction of LD as a whole, since if Bogaty and Gosain’s views were a community norm then certain argument types would simply die out. We should instead strive to find the most true arguments and vote for those in round; and this means we should allow every argument, even the ones we think are stupid.

Footnote

[1] http://nsdupdate.com/2016/03/16/an-alternative-view-of-offensedefense-by-michael-bogaty/

[2] Note that this is distinct from a similar view called epistemic modesty, which says that judges should not make absolute decisions about meta questions in debate but instead assign probabilities to each view. The classic example is that the epistemically modest judge might assign a 40% probability to utilitarianism and a 60% probability to deontology, but still vote for the utilitarian if the utilitarian harm is much larger than the deontological one. While epistemic modesty is possible under an offense-defense view, it is not a necessity and this article does not concern it.

[3] I realize that my use of the term “outcome” here might lead some to believe offense-defense is purely a consequentialist view. I say “outcome” due to lack of a better term, but one should essentially consider each relevant contention-level offensive argument as an “outcome” here. For example, the rights violation that a handgun ban might entail can be considered an “outcome.” The value of that outcome would be given by framework-level arguments.

[4] https://en.wikipedia.org/wiki/Von_Neumann–Morgenstern_utility_theorem

[5] To clarify, “utility” is simply a value assignment, so it doesn’t commit us to beliefs about the value of consequences or even commensurability of such values, which epistemic modesty would. For example, one could say that under a deontological framework death would have a utility of 0, while a promise broken has a utility of -1,000. If we cannot even comfortably assign number values operating within an ethical framework, there arise problems about evaluation when both sides have offense. But that is outside the scope of this article.

[6] I don’t want to get into the philosophical weeds on this issue, since of course I am no expert on philosophy and many people will likely disagree with me on this. But in the context of debate, at least, I think I can say with confidence that nobody has reached absolute certainty about any philosophical premises.

[7] I don’t think judges should completely abandon “tabula rasa” views and intervene on the side of what they believe to be true. Rather, what I mean by an open game is that we should not design rules (in this case, changing the meaning of terminal defense) that encourage or discourage certain argument types.