This post is based on: “The Slippery Slope Argument – Probability, Utility & Category Reappraisal,” written by Adam Corner, Ulrike Hahn, and Mike Oaksford and included in the 2006 Cognitive Science Conference Proceedings in the Cognitive Science Journal archive. The authors say that it is usually classified as a fallacy of reason, yet frequently used and widely accepted in applied domains such as politics, law and bioethics. They note that the slippery slope argument remains a controversial topic in the field of argumentation, and possesses the somewhat undignified status of “wrong but persuasive”. Having been a part of political and legal decisions, the slippery slope is ever present in my experience, although I never thought of it as a fallacy. Thinking about it though, it is what you tend to use when you think that you are going to lose the argument. I think it is interesting that every day people do not only use the technique, but they usually label it as a slippery slope which I guess shows what a powerful metaphor it is. I was also unaware that there is a “field of argumentation.”
The slippery slope is a consequentialist argument that warns against a particular course of action on the grounds that it will lead to an undesirable outcome. A slippery slope argument (SSA), however, posits not only a negative outcome, but the idea that this outcome might in the future be re-evaluated as positive, if an initial proposal goes ahead. The slippery slope has four distinct components:
1. An initial proposal (A)
2. An undesirable outcome (B)
3. The belief that allowing (A) will lead to a re-evaluation of (B) in the future
4. The rejection of (A) based on this belief
The alleged danger lurking on the slippery slope is, therefore, the fear that a presently unacceptable proposal (B) will in the future be re-evaluated as acceptable. If we withhold the right of free speech from a neo-Nazi organisation, what will prevent us from censoring legitimate political dissent in the future? The proponent of this argument is inherently appealing to the malleability of public opinion to reject an otherwise appealing course of action.
Corner et al state than an argument’s strength is a function of an individual’s initial level of belief in the claim, the availability and observation of confirmatory or countervailing evidence, and the existence and perceived strength of competing hypotheses. An individual’s belief in an argumentative claim can vary from 0 (no conviction) – 1 (total conviction), and is constantly being updated by (relevant) incoming information. The context, but more importantly the content of the argument and the beliefs of the arguer themselves will dictate the probabilistic value of an argumentative claim.
Using a Bayesian model of argument strength, an SSA is convincing to the extent that its consequences seem probable given the available evidence. In one sense, therefore, an SSA can be analyzed as a simple conditional probability – i.e. what is the chance of (B) occurring given
(A)? Consequently, one would expect SSAs whereby the initial proposal is likely to bring about the feared outcome to be stronger than ones where that probability is low. SSAs advocate decisions and as such are not just arguments about factual claims as most (so-called) fallacies are. One consequence of this is that we might even choose to avoid an action with only a low probability outcome, as long as that outcome is catastrophic enough. It should be the case that “the more probable the causal connection is, and the more we want to avoid (B), the stronger the argument”.
The researchers did two experiments, (Their paper includes the details.) and demonstrated empirically that SSAs vary predictably in their acceptability, and that this variation is broadly captured by a Bayesian account of argument strength. But do people have good reason to be persuaded by at least some slippery slope arguments. Does ‘slippage’ occur in the real world? The majority of the concepts that pervade our everyday argumentation are indeterminate. When advances in gene therapy are discussed, therefore, the specter of Nazi Eugenics is raised because the concept of pro-social genetic engineering is vague, and membership of the category “acceptable practice” is a dynamic and fluctuating process. Classification is heavily dependent on the set of instances to which the category label has been applied. There is a systematic relationship between the items that have been classified as belonging to a category and subsequent classification behavior. Encountering instances of the category at the category boundary will extend that boundary for subsequent classifications. According to the authors, there are numerous experimental demonstrations of so-called exemplar effects, that is, effects of exposure to particular instances and their consequences for subsequent classification behavior. For example, observing that a dog that weighs 10kg is considered underweight invites the conclusion that a dog that weighs 10.5kg is also underweight. With only the information that a 5kg dog is underweight, and a 15kg dog is overweight, however, one might not be so compelled to draw this conclusion.
There is then a feedback loop inherent in the classification of new data into an existing category, whereby that classification alters the category itself in a way that could naturally give rise to slippery slope arguments. So slippery slope arguments may be based on a category boundary extension process. This again shows the strength of the metaphor as the top of slopes are certainly boundaries.
I see a tie to fuzzy trace theory. When comparing adolescent behavior to adults, Valerie Reyna says that adults may just dismiss certain behaviors as unacceptable while teenagers weigh them more rationally. This is the situation whereby the outcome is improbable, but catastrophic so maybe adults see a slippery slope or at least a category of which they do not want to be part. In the decision landscape, there may be a certain steepness of slope that we will not stand next to based on our stage of development and individual factors.