Resisting Persuasion
9 December 2001
Francis
[Material in brackets is in need of reformulation or proposed omitted]

Vignette

The classic folktale of persuasion, which exists in several traditions, recounts the responses of a supposedly wise and sympathetic judge. Hearing the passionate testimony of the prosecution, he exclaims, "I believe you must be right!" After the defendant has delivered an equally brilliant defense, he again exclaims, "I believe you must be right!" His clerk politely whispers in his ear, "Your highness, surely they cannot both be right!" The judge turns to him and exclaims, "I believe you must be right!" (e.g., Idries Shah, who tells it about Mullah Nasrudin).

Introduction

Recent theories of persuasion [cite] have proposed a series of cognitive mechanisms or processes that to explain the phenomenon. Several areas have been examined, such as product advertising, public service announcements, religious rhetoric, and political propaganda. The analyses are insightful, convincing, and throw light on the detailed processes involved. Yet the very success of this enterprise threatens it with an abundance of riches, as in the story of the persuadable judge. If the cognive mechanisms that these theories propose are indeed effective, it follows that an appropriate implementation or use of these mechanisms in political cartoons and texts should reliably result in actual persuasion. Yet such a prediction clearly overshoots it goal. The phenomenon to be explained is the actual efficacy of persuasive strategies. It is well known that political propaganda, product advertising, and other forms of persuasion are not always effective.

Put differently, the paradox of research into cognitive models of persuasion is that it appears to be undermined by its own claims. In their choice of texts, advertisements, and cartoons, cognitive theorists select examples that they suggest employ mechanisms of framing and blending in effective ways. Curiously absent, however, is the claim that these examples actually persuade the authors themselves or their readers. Indeed, to claim that the examples are persuasive, in the sense of producing the desired change in attitude and intention, would clearly be absurd. Does not this fact in itself invalidate the claims put forward?

In the following, we present a model of how is it possible for us as readers or viewers of persuasive texts and images not to be convinced by them. The model is that of reverse engineering. In brief, we suggest that effective resistance to persuasion is both possible and commonly effected through a metacognitive operation of reconstructing the mental modeling undertaken by the person who produced the persuasive stimuli. In our view, this reconstruction needs to be fairly accurate to be effective; misguided reconstructions may produce inappropriate or ineffective resistance.

Our model of cognitive resistance does not invalidate existing theories of persuasion. In fact, we argue that effective persuasion is central to the phenomenon of resistance. It is because the cognitive processes involved in persuasion are routinely effective that resistance is necessary, and the design of effective resistance must closely mirror the design of effective persuasion.

Cognitive resistance

A commonsense model of cognitive resistance might suggest three dimensions to the phenomenon. We may resist information that contradicts our existing beliefs because the information is new and unfamiliar (the processing cost of integrating the new information is experienced as too high), because the information is unflattering (the emotional cost of adjusting to a less positive self image is experienced as too high), or because the information is false (the utility of integrating false information is frequently negative). More generally, we suggest that cognitive resistance functions to protect the inidividual's investment in a particular network of beliefs and assumptions. From a cognitive perspective, the question is what strategies are utilized to achieve this goal. We may wish to include in the arsenal of resistance strategies such as selective attention, cognitive blocking, distraction, and obfuscation. What is common to such strategies of interrupted processing is that the cognitive processes invoked by persuasion theorists are not in fact evoked. Consequently, these forms of resistence do not present an immediate theoretical difficulty to cognitive accounts of persuasion. To account for strategies of interrupted processing to achieve effective resistance opens a different set of issues that we will not pursue further.

More interesting for our present purposes are those cases of resistance where the individual subject to the persuasive message (the target of persuasion, or ToP) in fact evokes all the relevant cognitive processes that persuasion theorists propose enable effective persuasion, but nevertheless succeeds in mounting an effective resistance. This, we suggest, is the typical situation in theoretical work on persuastion, where the both the writers and the readers must necessarily evoke the relevant cognitive processes that are posited as the effective means through which persuasion operates. If these processes are not successfully evoked, but avoided through some variant of interrupted processing, the analysis of the persuasive text or image at hand must ipso facto be inadequate. The puzzle we wish to address, then, is how there can be a set of mental processes that on the one hand is (correctly, in our view) hypothesized to cause persuasion, yet on the other that the appropriate and circumstantially relevant activation of these processes nevertheless fails to produce the predicted result. The solution we propose to this puzzle is the cognitive operation of reverse engineering. Before we get to that, however, we need to sketch out a model of persuasion.

Persuasion

Persuasion means that one person is attempting to persuade another of something. That is to say, the concept of persuasion implies that the person attempting the persuasion has produced two mental spaces: one in which the other person has disposition A, and the other in which he or she has disposition B. The task to be undertaken by the persuader is thus to produce a stimulus that will cause the target of persuasion to change from the actual disposition A to the desired disposition B. In the case of mass communication, this task is typically formulated statistically. Given a population in which many individuals have disposition A, the task is to generate a stimulus that has a certain likelihood of causing individuals in the target audience to shift their disposition from A to B.

Generating the persuasive stimulus

The act of generating a persuasive stimulus, we would like to suggest, crucially involves two steps: audience modeling and exploratory blending. Both of these processes rely on conceptual integration.

In the first phase, audience modeling, the persuader puts himself in the place of his target audience. He does this, we suggest, by temporarily adopting what he believes to be the beliefs and dispositions of the individuals he is attempting to persuade. The reason such a simulation is potentially effective is that the persuader shares a large set of cognitive processes with his target audience -- processes the diagram locates in the generic space:

[Figure 1: Generic space: common inference systems; first input space: self; second input space: target audience; blend: persuaded self-as-other].

The degree of precision of the content of the second input space may vary. If the target audience is both well known and well defined, it is possible for the persuader to construct a self-as-other blend of great specificity. In most cases of mass communication, the degree of precision is modest, as the target audience is diverse. [Possible example: the use of focus groups in presidential campaigns; Bush Sr. found the backdrop of a flag very effective with varied audiences.] The reliance on the generic space must correspondingly increase.

We suggest this operation has the properties of a simulation in the sense that it is designed to access the relevant inference systems of the other, in order to be able to predict and thus to manipulate the products generated by these inference systems. The resulting blend described by this figure may thus be compared to an operating system. The blend itself does not produce any interesting results; rather, it is a platform from which the second stage of the process of generating a persuasion stimulus can begin.

In the second stage, the persuader explores a vast potential possibility space of persuasive stimuli. This space is so large that persuaders -- cartoonists, advertizers, political handlers -- typically adopt a range of strategies to limit which possibilities to explore.

The area defined may be constrained by factors such as topical events and common knowledge. [Possible example: political cartoons around Thanksgiving tend to favor the inclusion of turkeys and social rituals associated with the holiday event.] [An example of a very broadly targeted campaign: the movie Triumph of the Will.]

Just as importantly, the areas explored are constrained by what constitutes effective cognitive blends. [We should discuss these constraints in greater detail; they clearly involve the selection of appropriate of frames.]

In order to be effective, a persuasive stimulus must pass a series of tests. [We should discuss what these tests are; I have in mind things like novelty, humor, and topicality.]

In brief, in order to generate a suitable persuasive stimulus, you must place yourself in the position of the person or persons to be persuaded and come up with a new and attention-grabbing conceptual blend. Various candidate conceptual blends -- persuasion stimuli -- are then tried out within the structure of the simulation, until one that is appropriate and effective is located. The reason that a strategy of modeling the other -- of pretending to be the other -- works is that the persuader and his target audience shares a complexly structure generic space, as well as some degree of specificity with regard to the other's beliefs and dispositions.

The blend has some probability of achieving the goal of persuading the target audience; we turn to this next.

Becoming persuaded

[Add your cognitive account of persuasion: how does it work?]

Reverse engineering

[Spell out the model -- in brief, the idea is that effective resistance is possible and in fact commonly implemented by attempting to "see through" the persuasion stimulus. This involves recreating the potentially manipulative processes intended by the persuader and built into the persuasion stimulus. We should pick some examples and show how this may be done; it would also be useful to run some experiments and see if this theory is supported.]

Resisting persuasion

The astute reader will no doubt long since have noted the irony of our resolution to the problem we have placed before us. Persuasion theorists, we argued, have produced detailed and in our view accurate models of the cognitive processes responsible for persuasion. Yet the examples they use in their own articles, chosen for their preeminent effectiveness, do not in fact persuade. Our present theory amounts at first blush to a suggestion that such examples fail to persuade precisely because and to the extent that they have been accurately analysed. That is to say, we may appear to be saying that the type of analysis performed by cognitive theoriests is precisely the analysis performed by a typical subject of persuasion when he or she mounts a successful cognitive resistance. In both cases, there is a reverse engineering of the mental processes performed by the originator of the persuasion stimuli.

[We may want to do some empirical work here -- or refer to work done for the previous section -- to determine the distinctions between the proposed reverse engineering performed by naive subjects and the cognitive analysis performed by cognitive theorists. The general idea would be that the conceptual primitives likely differ, perhaps considerably.]

[Add a discussion on the issue of whether the reconstruction must be accurate or not in order to produce an effective resistance. The arguments here are twofold.

First, it is possible to misattribute a series of mental states to the originator of the persuasive stimuli (OOPS), for instance by attributing an intention to deceive, harm, and misinform to an honest and well-meaning communicator. This would nevertheless result in an effective blocking of the message.

Second, it is possible to succeed partially in a reconstruction, but to fail in some detail. In that case, would we predict that the missed detail would be persuasive? It would seem so.

Finally, then, by making an accurate reconstruction a requirement for accurate resistance, we may be situating cognitive resistance within a larger framework of communication, where communication is understood in terms of producing accurate inferences concerning the content of another's mind -- that is to say, something close to relevance theory.]

Conclusion

In this article, we have proposed that ...

Returning to the story of the gullible judge, we suggest that his amusing naïveté consists precisely in his failure to reconstruct the mental states that gave rise to the conflicting testimonies of each party in the conflict. [His failure to accomplish this task renders him subject to effective persuasion in a ridiculously exaggerated manner.] We will spare the reader a further analysis, as a meticulous reverse engineering of humor unfortunately tends to produce an effective resistance, inadvertenly spoiling the joke.