This paper proposes a pragmatic analysis of two so-called fallacies in argumentation, namely the 'Slippery Slope' and 'Guilt by Association'. I will examine their rational use, and argue that they exemplify at least partially non-cooperative, but still inference-based conversational moves.
Pragmatic theories of the (neo|post)Gricean type typically assume that conversation and inferences can be modeled by the (rational) interaction of a speaker and a hearer, and also, that pragmatic inferences are based somehow on a cooperative interaction between speaker and hearer.
Both of these assumption have been challenged.
So-called ‘argumentative’ theories of pragmatics
1And while it is not often developed, the audience in itself can overlap to various degrees with the DECIDERS.
2Levinson (1988) or Clark (1996) have proposed to enrich the speaker-hearer dichotomy in other ways. Their move is motivated by issues such as turn-taking and participation. I — like the literature on argumentation — focus however on strategic interaction. As far as I see, these proposals are perfectly orthogonal.
Gricean pragmatics operates within the assumption that speaker and hearer are cooperative, as embodied by the cooperative principle. While rarely stated explicitly (but see Fox (2014) on this issue), one implication one can draw from this idea is that in non-cooperative contexts, there should not be any pragmatic inference. The resulting ‘Kumbaya’pragmatics may not correspond to Grice’s intentions, but it makes certain empirical predictions. These predictions, however, rely on the fact that the participants the speaker is cooperative with are correctly identified.
Argumentative communication provides a
testing ground for the idea that pragmatic inferences
necessarily necessitate cooperative interaction, or
necessarily involve a common ground
Now, the idea that argumentative communication is based on inference is not exactly new; it can be traced back at least to Aristotle. He noted that ‘normal’ argumentation does not contain comprehensive and logically valid demonstrations, but rather abbreviated versions of it. Aristotle points out — in an explanation that has a Gricean ring to it — that generally, speakers in argumentation do not use full syllogisms, but rather the shorter enthymemes:3
The Enthymeme must consist of few propositions, fewer often than those which make up the normal syllogism. For if any of these propositions is a familiar fact, there is no need even to mention it; the hearer adds it himself. [. . . ] we must not carry its reasoning too far back, or the length of our argument will cause obscurity: nor must we put in all the steps that lead to our conclusion, or we shall waste words in saying what is manifest. 4 Avoid obscurity and be brief can be included in a cooperative setup (as in Grice’s maxims of manner), or they can be seen as self-interested, rational ways of dealing with the (limited) attention span of the audience. But crucially, these maneuvers entail making use of inference by the audience. While Aristotle seems to consider here essentially inferences from the common ground, this is not always the case. Consider “Make America great again. Vote Trump.” Assuming this to be an instance of argumentative communication, this relies (at least) on the unstated proposition “Donald Trump can make America great again”. This proposition is arguably not part of the common ground of American voters.
My intention here is to follow up on remarks by Volokh (2003), namely that mechanisms that are generally considered by logicians to be fallacies (for instance, ad hominem arguments) are better conceived of as heuristics for real-time decisionmaking by rationnally ignorant agents.5
I will also assume that there are principles of
rational argumentation, such as giving the strongest
argument at the arguer’s disposal
More precisely, I will show with the examples of the Slippery Slope and ‘Guilt by Association’ that these are contexts where the notions of audience 3A enthymeme is often considered to be a truncated syllogism — that is, a syllogism where one premise is lacking.
4Aristotle, Rhetoric, 1357a 16, 1395b 25, as cited in Hamblin (1970, 71); my emphasis.
5Rational ignorance means that we often have to make decisions while ignoring their precise outcome. and opponent need to be kept apart, but where it is doubtful that we face a fully cooperative setting with either of these. 2
The argument of the slippery slope involves advancing an argument against some proposition A based not on the intrinsic merits or deficiences of A, but rather on the assumption that once A is in place, there would be no way of meaningfully opposing B, which is assumed by the arguer to be undesirable. Arguments of this kind often appear in discussions concerning gay marriage (A), which is opposed not as such, but which is argued to lead to a state where one could not oppose further develo legalization of adoption by gay couples, or even the legalization of polygamy, incest, bestiality, etc. This argument is fallacious (that is, in need of additional inference), since there is no logical, entailment-based link between, e.g., polygamy and gay marriage.6
Volokh (2003) provides a very complete study of the mechanism of the slippery slope, and makes clear that at least in some cases, the risk of going down a slippery slope is real. While his paper contains many observations that are relevant to linguists, his main problem — as a legal scholar — is to identify how the slippery slope works in the real world, which are the mechanisms of slippage, and how it can be avoided. His article, however, does not address directly the issue as to when the argument of the slippery slope is rational or appropriate, which is the focus of the present paper.
Volokh identifies several mechanisms that can cause slippery slopes, of which several entail mixed motives (what Volokh calls “multi-peaked” preferences). An example Volokh gives is the proposal to install video surveillance in a town, where there are in principle three alternatives: i) oppose it, and remain in current state (note this 0); ii) vote for a version where cameras are not connected to facial-recognition software and tapes are rapidly destroyed (A); or iii) vote for a version where cameras are connected to facial-recognition software and tapes are kept for a long time (B). In a context where voters are not only motivated by concerns 6I take it that even the staunchest opponent of gay marriage would have to concede that there might be, in principle, a society which bans polygamy, but nevertheless allows gay marriage. I furthermore take it that the disagreement hinges on the question whether such a state of affairs is attainable and maintainable for the real world, given the current state. about privacy vs. security, but also by the financial cost of the system, some people will oppose video surveillance for cost-reasons, even if they are in principle favorable. Therefore, there may be no way of directly going to B from 0. However, if A is enacted, the cost motive for opposing B will be removed (since tapes and software are much less costly than the installation of the cameras in the first place), and in a subsequent vote, B could be adopted. Therefore, people with a preference profile of A > 0 > B should rationnally oppose the move to A, even though it is their preferred option, because they would end up with B, which they strongly oppose.
Let us come back to the issue of cooperativity. Volokh (2003, 1034f.) makes the following observation with respect to slippery slopes.
Slippery slopes may occur even when a principled distinction can be drawn between decisions A and B. The question shouldn’t be “Can we draw the line between A and B?”, but rather “Is it likely that other citizens, judges, and legislators will draw the line there?” [. . . ] Societies are composed of people who have different views, so one person or group of people may want to oppose A for fear of what others will do if A is accepted. And these others need not constitute a majority of society: slippery slopes can happen even if A will lead only a significant minority of voters to support B, if that minority is the swing vote.
According to Volokh, thus, a slippery slope will not occur if one can trust the deciders (the other citizens, judges, and legislators). Therefore, an argument of the slippery slope is a sign of lack of trust, and not of principled and uncompromising cooperation. In order to investigate when it is rational to use an argument of the slippery slope, I will try to make explicit the decision process in terms of conditional probabilities, along what has been done by Merin (1999).
First of all, the argument of the slippery slope is an indirect argument. Going down this route should only be done if the direct approach — that is, directly opposing A — does not appear to be viable. This in itself is a sign of a weak position, and it is rational only if — given the arguer’s information state, there is a sufficient majority of deciders in favor of A, such that the change towards A can be enacted. Furthermore, in order for the argument of the slippery slope to make sense, there are two further requirements: first, it must be the case that the probability of implementing B given A is higher then the probability of implementing B. This can be written as follows: P(Bji[A])7 > P(Bji). This is probably too weak a requirement, since it must be also the case that the the deciders are in majority opposed to B.
Second, the slippage towards B will only work as an argument if B is considered as sufficiently repulsive to motivate a rejection of A even if if A is the preferred option. Let us note the expected utility of some action or state S EU (S). One way of thinking about this is the following: The expected utility for state A, given the probability of slippage from A to B P(BjA), and the expected utility of B, will be the expected utility of A plus the probability of slippage multiplied by the expected utility of B. (1)
EU (A) + P(BjA)
EU (B) If (1) is negative, a rational agent should reject a move to A. When will this be the case? The lower the (positive) expected utility of A, and the higher the risk of slippage and the (negative) expected utility of B, the stronger the trend to rejection.
Therefore, the slippery slope will be most appropriate if A is too popular to be attacked directly, if B is as repulsive as possible, and if, at the same time the risk of slippage from A to B is considered to be high among the audience. It seems obvious that in most circumstances, these conditions will not be met — especially if the passage from A to B is under full control of the audience, and is not impinged on by issues of applicability (possibly under the control of third parties — like the justice or the police). Hence, if the audience are the deciders, the mere suggestion of the possibility of a slippage can be interpreted as a vote of non-confidence towards (at least) a majority of the deciders. Therefore, the argument of the slippery slope can be detrimental to the arguer and his thesis.
Notice, though, that the argument of the slippery slope can only work if there is a justifiable lack of trust with respect to the deciders, and that it is therefore not an argument that is built on unconditional cooperation. 3
Guilt by association is an argumentative move where the opponent’s position is rejected based on the assertion that this position was also held by other, less-than-recommendable people (noted henceforth as bogeyman). For instance, reductio ad Hitlerum is an instance of guilt by association, but it englobes also red-baiting on the other end of the 7I note as i[A] the information state i augmented with A — which may cause changes other than merely adding A. political spectrum. Once again, this move is classified as a fallacy, since the fact (or still less, the assertion) that, say, Hitler held some view (against smoking, or for vegetarianism, for instance) cannot generally be taken as a reason for dismissing this view without additional arguments (or contextual inference, for that matter). More precisely, guilt by association depends on a relevance-implicature.
Now, when is such an argumentative move rational? Notice that guilt by association may have two different aims: first, if the opponent is (part of) the audience, the opponent is to be shamed into accepting the arguer’s view. Second, if the opponent is not part of the audience, it aims to exclude the opponent’s arguments from consideration by the audience.
Let us start by considering the first strategy. The wished-for reaction in the opponent would be the following: The arguer asserts that only bogeyman would hold opinion f .8I asserted that position, but I do not wish to be identified as bogeyman; therefore, I (possibly publicly) abandon my position, and adopt the position of my opponent.
In the second scenario, where the opponent is not a member of the audience, the basic process is like above — but since members of the audience have not brought forward any claim, they will not have to publicly retract.
Generally, the move is based on social exclusion, and is intended to remove the opponent from the people that are entitled to present counterarguments with respect to some theme to the audience. Therefore, the audience should assume that the proportion of people holding the opinion is low (whether this is true or not is another question) — since otherwise, it will not provide a good means of social stigmatization. If the stigmatized opinion is widespread, and even if the bogeyman as such is strongly rejected, guilt by association may backfire, and provoke rejection towards the arguer.
The latter thought process can be explicited as follows: I hold opinion f , and I know that I am not bogeyman. Furthermore, I know that a considerable part of the audience hold opinion f , and 8This may appear as a unnecessary strengthening of an argument of guilt by association, which is likely to go rather like “bogeyman thought|said that, too”. However, even the scariest bogeyman will have countless opinions that are perfectly mainstream in the considered community, such as “It is right to drive on the right hand side”, etc. Therefore, for the argument even to be relevant, it has to be the case that that particular kind of opinion must have some link to what makes that bogeyman a bogeyman. are no bogeymen. Therefore, the argument of the arguer does not hold. The arguer could have presented another type of argument, but he chose this one. Therefore, this must be what he thinks to be his strongest argument at that point. Since it is not correct, the case must be dismissed.
Guilt by association operates with a strongly negative social emotion (which is the prototypical noncooperative move), and pits that against whatever evidence the opponent has for his position. If the opponent thinks that that evidence holds up well, or feels strongly against being publicly shamed, guilt by association will fail.
Finally, there is an intrinsic problem with guilt by association arguments: the more evil bogeyman in the opinion of the audience,9 the stronger the argument. And as the bogeyman is evil, this implies that the opponent’s argument should not even be acknowledged. However, if the audience identifies this as the intention behind the use of the guilt by association argument — and if (at least a considerable part of) the audience holds that opinion — it will be received as a refusal to discuss that particular issue, and communication may break down. In any case, guilt by association is a highly polarizing type of argument, whose aim is rather to mobilize the own camp within the audience than to bring around people holding the opponent’s view. So, if the audience and the opponent are identical (or if the opponent is a critical part of the audience), guilt by association should in most cases be avoided. 4
I have tried to show that in argumentative discourse, the familiar speaker-hearer dichotomy is too simple to meaningfully describe the rational interaction of discourse participants. I also tried to show that there are inference-based discursive moves which are clearly non-cooperative, not only with respect to the opponent, but also with respect to parts of the audience. In case of an argument of the slippery slope, the basic outline of the argument is based on the idea that the deciders cannot be sufficiently trusted not to go down the slippery slope. In case of guilt by association, I have argued that the argument is based on (the threat of) social ostracism, and therefore equally non-cooperative.
9In an assembly of neonazis, the reductio ad Hitlerum would obviously not qualify as an argument of guilt by association, but rather as an argument by authority. I would like to thank the three anonymous reviewers for their comments and suggestions on a previous version of this paper. All remaining errors are mine alone.