The Prospects for Neuro-Exceptionalism: Transparent Lies, Naked Minds

2008. The American Journal of Bioethics 8(1):3

This Article does not have an abstract.

Introduction

The conviction that neuroscience and neurotechnology pose new ethical challenges, particularly concerning privacy, appears to rest on the belief that they yield a new, more intimate or personal kind of information. This belief in "neuro-exceptionalism" may seem more plausible than the belief in genetic exceptionalism that influenced the public debate on the uses of genetic technology. But an examination of some possible applications of neurotechnology - identifying behavioral dispositions, detecting lies, reading minds - raises questions about the plausibility of neuro-exceptionalism, in light of present developments as well as future prospects.

In dozens if not hundreds of science fiction movies and stories, a machine is used at some point to determine what someone is really thinking. Does she really remember that terrible event? Is she really telling the truth? Does she really love her husband? What was she really thinking the moment of the crime? The hold of this theme on the popular imagination quite likely underlies some of the current anxiety about neuroscience. Will it provide us with a significantly different kind of information about people's thoughts and feelings, like those science-fiction machines? Will it yield accurate and reliable conclusions about what is on, or in, a person's mind, regardless of her intention or willingness to share that information? Will it pose a special threat to privacy?

These concerns are provoked by actual developments in neuroscience and technology, particularly in brain imaging. We can now see what people's brains "look like" when people are performing various cognitive functions and displaying various emotions, and we can observe striking differences in the brain images of people differing in age, gender, and psychiatric diagnosis. Yet, while these images may seem to reveal much that was previously concealed, some of the questions we just raised are vague and unclear. What do we mean by "significantly different kind of information"? Is that not a matter of degree? What constitutes a "special threat to privacy"? Does that not that depend on cultural expectations and current protections? In this article, I will suggest how we should address these questions and sketch some answers.

THE IDEA OF EXCEPTIONALISM

In order to address these questions, we would do well to recall discussions of a similar set of issues in genetics, particularly the dispute over so-called genetic exceptionalism. The central issue of that dispute was whether genetic information is significantly different from other medical information. Some held that genetic information reveals details that are more intimate, sensitive, and predictive than other medical information. Moreover, because medical information is generally thought to warrant privacy protection, the more sensitive or "private" nature of genetic information means that its acquisition and dissemination would require greater or special safeguards.

These claims have not held up well to critical scrutiny. Several commentators have persuasively argued that what genetic information can reveal - about ancestry, the risk of having or acquiring a particular disease, or a stigmatizing condition - is no different from the sort of information non-genetic medical examination can reveal (Murray 1999, Botkin and Green 2003). And if genetic information is no different, it can hardly be, in general, more intimate or sensitive. Genetic information, then, is not exceptional; it does not raise novel or special privacy concerns.

Two caveats are important to note: First of all, even if there is no fundamental difference between genetic and non-genetic medical information as far as what it can be reveal, the perception that a particular piece of information is genetic - e.g., acquired from a genetic test - might give it a greater significance because of a general belief that genetic information reveals matters that are inevitable or "essential" to the individual. But such perceptions do not support genetic exceptionalism, because it is not a descriptive thesis. It claims that such perceptions are based on the biological facts.

Secondly, genetic exceptionalism is about the content of the information, not about how it is acquired. According to genetic exceptionalism, there is a fundamental difference between genetic and non-genetic medical information with respect to their personal, intimate, or private character - genetic information is necessarily more private or revealing than non-genetic medical information. Nevertheless, we should note that a threat to privacy can arise from the acquisition of a new kind of information or from a new way of acquiring a familiar type of information. (This latter was the threat posed by electronic surveillance.) Genetic technology can present means of acquiring information that can raise distinct privacy concerns - e.g., with regard to using genetic technology to obtain information from oral swabs or shed hair without the subject's knowledge or consent. Thus, even if genetic exceptionalism is false and so there are no special concerns about privacy raised by the possession of genetic information, special concerns can still arise over new methods - including genetic methods - of acquiring familiar information.

Even with these qualifications in mind, rejecting genetic exceptionalism has the salutary effect not only of avoiding bad policy but of demystifying genetic information. In denying that such information is exceptional, we recognize it as a form of medical information, which may be no more sensitive, or no less uncertain and ambiguous, than more familiar forms. The only exceptional danger posed by genetic information arises from the widespread, recalcitrant belief that it is exceptional.

Can we say something similar about "neuro-exceptionalism"? The answer to this rests on how we characterize "neuro-information" and with what other types of information we compare it. In the case of genetics, the type of the information is clear: we are discovering the presence (or absence) of particular genes or alleles whose significance is specified by their causal properties. Neuroscience, however, presents us with a more complex picture: it reveals structures and activities that do not always bear a causal relationship to the behavior or mental states we seek to predict or explain. Rather than attempt a general answer about information-type, I will consider examples of three types of information that neuroscience might yield - two tied to current research, one clearly in the realm of science fiction - and to assess whether a case for neuro-exceptionalism could be made for any of them, whether any of these types of information would be fundamentally different from other information already available. As I will try to show, the claim of neuro-exceptionalism is doubtful in the first two cases, but seems warranted in the third. The third, however, is and will likely always remain science fiction.

NEURONS, GENES, AND CAUSES

Suppose that the neuroscientific information in question concerned the presence or absence of certain chemicals, structures, or activities in the brain, the significance of which was specified by their causal properties. For example, suppose we found low levels of the neurotransmitter serotonin in an individual's brain or found that the serotonin being released was reabsorbed too quickly and completely. Such information, we are told, would provide evidence that the individual has a propensity or disposition to depression (cf. Kandel et al. 2000). Is this type of information fundamentally different from, for example, genetic information? Consider the case of the geneticist who discovers that an individual has the short allele of the 5-HHT gene. It would seem that she can draw the same kind of conclusions about a propensity towards depression since these particular alleles are the usual causal factors in the development of excessive serotonin reabsorption.¹ Because the geneticist has simply identified an earlier point on the causal path, there does not seem to be a fundamental difference in causal significance between this type of neuro-information and genetic information. To be sure, the genetic information might differ in its evidentiary weight - because the genetic information concerns an earlier point in the causal path, we might claim that the association of genes with a certain behavioral disposition was of lower probability that the association of neurochemicals and structures with that disposition. But that difference in degree does not seem to warrant claiming a fundamental difference in ethical significance. How could information regarding an individual's serotonin reabsorption be considered a substantially different or greater threat to privacy than information regarding her having the 5-HTT allele?

When we consider the causal significance neuro-information, the burden of proof is on the neuro-exceptionalist to explain and argue for a fundamental difference. One way to discharge that burden would be to argue for a "neuro-determinism": although the causal association between genes and behavioral dispositions is only probabilistic, the presence of neuro-chemicals, structures, etc., causally determines the behavioral dispositions. Arguably, that would constitute a difference with significant ethical implications. Nevertheless, such a position is untenable. People reject genetic determinism because they rightly acknowledge the importance of environmental factors in the development and expression of behavioral dispositions. The same considerations that undermine genetic determinism also undermine neuro-determinism.

A consequence of rejecting this type of neuro-exceptionalism is particularly noteworthy. More than a decade ago, a great deal of scholarly interest was directed at the ethical implications of behavioral genetics (Wasserman and Wachbroit 2001; Parens et al. 2005). If the preceding discussion is correct, then the ethical issues raised by this type of neuro-information - information regarding neurochemistry, structures, etc. - are not much different than those raised by behavioral genetics. Similar concerns about privacy, identity, social control, enhancement, and so forth, arise in both cases. Of course, this is not to deny that there might be a significant social or cultural difference between neuro-information and genetic information. And such differences can lead to different ethical concerns. But, as we noted earlier, the rejection of genetic exceptionalism - the rejection of a fundamental difference between genetic information and medical information - does not imply that there is not significant difference in the social perception of genetic and other medical information. The significance of rejecting exceptionalism, genetic or neuro-, is that it challenges the basis such social perceptions.

In sum, insofar as neuroscience merely identifies chemicals, neural structures, and their causal properties, neuro-exceptionalism is false - from an ethical point of view, there is no fundamental difference between this type of neuro-information and genetic information. None of this, however, should be taken to deny or diminish the important epistemological and scientific differences that exist between them. Nevertheless, new and exciting scientific discoveries do not always mean new ethical issues. And so it would seem in this case.

EVENTS, LIES, AND ASSOCIATIONS

The significance of genetic information, at least from an ethical point of view, lies in its causal claims - about what we can predict, what we can control, and what we can causally explain. If the same were true about neuro-information, then our discussion of neuro-exceptionalism could perhaps stop here. Interestingly, however, much of the information we can expect neuroscience to yield is not about causation.

Consider neurological information about events in the brain - e.g., blood flow, electrical activity, or other events that can be captured by neuroimaging technology. The significance of such information, in contrast to genetic information, consists not in causal claims but rather in the discovery of what the individual is currently thinking or feeling. That discovery need not rest on the assumption that thoughts and feelings are identical to specific neural images, or on a "reduction" of the former to the latter, or on a claim that one causes the other, but on empirically discerned associations such as between a pattern of neural images and a certain type of mental activity. A good example of this effort has been the investigation of neural images to detect lying. (For a recent review of this research, see Keckler 2006.) Does this type of neuro-information qualify for exceptional status? That is to say, is there a fundamental difference between this type of neuro-information - the information from neural images or other neuro-detection of lying technologies - and the information yielded by more familiar or conventional efforts at detecting lies?

To begin with we should note that brain events are hardly the only biological events that can be associated with what an individual is currently thinking or feeling. It should be plain that much of human communication turns on the recognition of the somatic accompaniments of what is being said. Tone of voice, facial expressions, gestures, and movements help to identify whether a particular speaker is feeling serious, humorous, sarcastic, fearful, etc. Indeed, polygraphy, the current technology of lie detection, is based on such associations.

Human beings are modestly good at detecting lies, even if you can fool all of the people some of the time and some of the people all of the time. If humans were bad at detecting lies in general, it would be difficult to understand how they could have survived as a species. In some cases we detect a lie based on the content of the statement: we know that the speaker does not believe what he is saying and that he does not realize that we know. In other cases we detect or at least suspect a lie based on the presence of physical symptoms associated with lying. Is the speaker nervous, is he perspiring, is he avoiding eye-contact, is he acting as if he is "making it up as he goes along"? We do not explicitly run down a checklist; we typically recognize the symptoms of lying implicitly, even if it takes some reflection to articulate what those symptoms are. Good liars are people who can suppress these symptoms.

In essence, polygraphs can be seen as a technology to amplify some of these symptoms. Rather than wait to see if there are beads of sweat on the speaker's forehead, we can determine whether the electrical resistance of the speaker's skin is changing (galvanic skin response), which typically precedes visible sweating. Rather than see if the speaker is acting nervous, we can determine whether his heart rate and respiration have changed significantly, often indications of nervousness. By amplifying some of the physical symptoms of lying, polygraphs can make it easier to detect. But, of course, they are not foolproof. Some people can learn to suppress even such amplified symptoms of lying. More importantly, the technology consists in the amplification of imperfect associations: it is not based on a biological analysis or a reductionist understanding of lying. The current technology of lie detection does not assume that there is a biological mechanism for lying or that all lies have some biological features in common. It is quite plausible to hold that lying is a heterogeneous category with vague borders shaped by culture - for example, to the extent that lying involves a breach of trust, its occurrence will depend on the cultural norms and practices that establish trust. A biological classification could not possibly capture the diverse, culturally determined collection of behaviors we call lying. But acknowledging this complexity no more undermines or challenges polygraphs than it undermines or challenges our ordinary abilities to detect lies. That is because polygraphs themselves do not detect lies; they simply amplify the signals we use to detect lies.

Because polygraphs are based on the amplification of ordinary but imperfect associations, though, they have two important shortcomings. First of all, we can expect polygraphs to yield false negatives and false positives - they will fail to identify some lies as lies and they will identify some non-lies as lies. But this is how it is with our ordinary detection of lying. The use of the behavioral and physical symptoms of lying is a really just a heuristic, a rule of thumb on which we must be careful not to place too much reliance. Second, by amplifying these symptoms, polygraphs risk introducing noise and distortion. For example, consider a situation where nervousness on the part of the speaker would raise the suspicion that he was lying. What if the speaker does not appear nervous in that situation, but a polygraph detects subtle signs of nervousness, e.g., elevated heart and pulse rates. Do such subtle signs, not rising to the level of ordinary symptoms have as much evidentiary force? Or are they likely to reflect random or independently caused physiological fluctuations of no evidentiary value? Although skilled polygraph users have developed various ways of addressing these shortcomings, the reliability of polygraphs remains a hotly debated issue (National Research Council 2003).

The neuro-detection of lying, however, looks at a broader ranges of phenomena than polygraphs; if it did not, it could hardly claim to provide a significantly different kind of information regarding deception. Polygraphy focuses on the affective symptoms associated with lying - fear, nervousness, guilt - whereas neural imaging focuses on the cortical regions of the brain and the cognitive symptoms associated with lying.² (Neurological lie-detection, like other neurological investigations, might also look at the activity of the amgydala - the "seat of emotion," but most lie-detection studies to date have had a cortical focus.) For example, fabricating a story often requires more cognitive activity than truth-telling, and some studies have indeed detected such increased activity in the context of lying (Spence 2001, Nuñez 2005).

Even though neurotechnology goes beyond polygraphy in the information it uses to detect deception, does it go beyond the information underlying common sense lie detection? We all recognize hesitation, halting, and equivocation as symptoms of lying, and skilled interrogators are doubtless trained to detect and amplify these symptoms. Neurotechnology seeks to detect this same type of information at an earlier stage, catching the brain in the time-consuming process of suppressing a truthful response or inventing a false one. But the association of lying with the cognitive activities actually detected by neuroimaging is still imperfect, subject to all the cautions and qualifications we mentioned earlier. The clarity of the visual images it yields has little or no connection with their accuracy as symptoms. If the neuro-detection of lying simply amplified the cognitive phenomena we already and ordinarily associate (imperfectly) with lying, then the differences in the information is more a matter of degree than kind. Neuro-exceptionalism would therefore be false for this type of neuro-information. The ethical issues raised by the neuro-detection of lying would seem not to be fundamentally different from those raised by polygraphy.

But is this the only type of information the neuro-detection of lying could yield? Could neuroscience discover a new kind of evidence for lying, one that does not rely on already recognized, imperfect biological symptoms? The functional cognitive modeling neuroscientists use to analyze deception into potentially detectable phases - e.g., truth-suppression or fabrication phases - is unlikely to yield such new evidence, since those models draw on, and are correctable by, common sense understanding of what lying is and what cognitive steps it requires (Spence et al. 2001). Some have claimed that "there are fundamental differences between deception and truth-telling at the neurological level " (Wolpe et al. 2005, 43). But given the widespread recognition of how little we know about the brain's workings, this claim is difficult to accept without a great deal more evidence and theory than we currently have. If the claim is just an instance of the general thesis that different mental states differ at the neurological level, then it is little more than a rejection of an extreme, implausible form of mind-brain dualism. We can just as legitimately say that there are fundamental differences at the neurological level between thinking about Paris and thinking about Rome. No conclusions follow about the possibility of a new kind of evidence for thinking about European cities. However, if the claim amounts to the much stronger thesis that a neuroscientific reduction of lying is possible, or that lying has a distinctive neurological signature, then argument and evidence are needed. The variation, vagueness, and cultural dependence in what we call a lie makes this seems unlikely.

Nevertheless, neuroscience might discover a new correlation or association between (some types of) lying and specific brain activities. Suppose, to take a completely fictional case, an association was discovered between lying and activity in the region of the brain associated with the olfactory system. This would provide a new kind of evidence for lying, which did not reflect any ordinarily recognizable symptom of lying. It could therefore form the basis of a defensible claim of neuro-exceptionalism. But as recent reviews of current research indicate (Illes 2004, Wolpe et al. 2005, Keckler 2006), such a discovery does not seem on the horizon.

THE SPECTER OF MIND READING

Finally, suppose the neuroscientific information consisted of the thoughts themselves - that obtaining information amounted to "mind reading" - so that the person's thoughts become uncontrollably transparent (or better, naked) to an observer. This is the deepest fear about the information neurotechnology could yield. Just as an unauthorized gaze on a naked body is a paradigm of invaded privacy, so an unauthorized gaze on a naked mind would amount to a completely new kind of information and so would constitute a special threat to privacy. Nevertheless, we have now entered the realm of science fiction; no one has suggested that we can get such information using neurotechnology. Indeed, it is not even clear what the information displayed by a naked mind would consist of. Would it be images, sounds, or words? Would it be a literary "stream of consciousness," like that described by William James or employed by James Joyce? Would it be like the cacophonous data stream in a multi-carrier transmission cable? We have no idea.

The actual data would not of course consist of words, etc., but rather of neurophysiological data - brain states - interpreted the correct way.³ Consequently, the naked-mind scenario would require the development of, in effect, a translation manual, connecting particular brain states or neural images with specific thoughts. What would such a translation manual look like? We immediately face the problem of what is called "the opacity of propositional attitudes" - even if two propositions are equivalent, the thoughts about those propositions might not be. Because the thought that Augustus was the first Roman emperor is different from the thought that Octavius was the first Roman emperor - even though Augustus and Octavius were the same person - the corresponding brain states would have to be identifiably different and so noted in the translation manual. Furthermore, the structure of the translation manual would be rather restrictive. For example, because people do not always believe the logical consequences of their beliefs, the brain state corresponding to having the belief that A and B would be different from the brain state corresponding to having the belief that A and having the belief that B. Indeed, it seems unlikely that the brain state corresponding to the belief that B could be a recognizable component of the brain state corresponding to the belief that A and B. Otherwise we would have a new way of doing mathematics simply by examining the brain state of a mathematician: we could determine whether one mathematical proposition entailed another by determining whether the brain state corresponding to a belief in the second proposition was a component of the brain state corresponding to a belief in the first proposition. In general, therefore, we cannot regard a brain state as being composed of recognizable component brain states that mirror the structure of the belief associated with that brain state. This means that the translation manual would have little or no recursive or generative structure: every possible belief would be a separate entry since no rule or algorithm could yield a new entry from previous ones. (It would be like a grammar book without rules but rather a list of all grammatical sentences.) Similar remarks would hold for desires, hopes, wishes, etc., insofar as they have a propositional content.

There would be further problems with a naked-mind translation manual. Consider the thought that Augustus was the first Roman emperor as held by a sexagenarian professor of ancient history and as held by an adolescent student. Even though the sentence, "Augustus was the first Roman Emperor" would mean the same coming from either mouth, we would expect the corresponding thoughts - what Frege, the founder of modern logic, called "the associated ideas" - to be different. Not only would we therefore expect their brain states to be different, we could not suppose that there is a recognizable component brain state that the professor and the student shared. It seems that we would need a separate translation manual for each person. Imagine the evidence needed to construct idiosyncratic translation manuals. In sum, even if we allow the theoretical possibility of a neuroscientific reduction of beliefs, hopes, etc. - which is clearly controversial - it would be practically useless for individual mind-reading.

It should be noted that this argument is directed at the feasibility of using neuro-information to identify the contents of a thought. It is silent about the possibility of using neuro-information to identify the kind of thought: it could well be that similar things are going on in the professor's brain and the student's brain when they are both afraid or depressed. Indeed, as we noted earlier, research has indicated that such is probably true for certain cases of depression. Nevertheless, such information does not identify the propositional content of belief, fear, depression, etc.

CONCLUSION

Not every new science or technology raises new ethical issues. Nevertheless, the excitement and enthusiasm that comes with new discoveries and developments is often accompanied by doubts and concerns over their ethical implications. Are these familiar concerns dressed in new skins or do we face new ethical challenges, calling for new analyses, frameworks, and principles? One way of addressing this question has been to ask whether the information from a new science or technology is fundamentally different, from an ethical point of view, than the sort of information already available. Does it raise new ethical issues or does it merely present new, perhaps more acute or urgent, instances of previously examined ethical issues? This is the point of an inquiry into exceptionalism.

With regard to neuro-exceptionalism, we have tried to show that the matter falls into three cases: Insofar as the significance of neuro-information lies in its claims about causation - what we can predict, what we can control, what we can causally explain - then this type of information is not fundamentally different from the information obtained in behavioral genetics. Things are different when we instead consider neuro-information whose significance lies in its identification of types of mental states. If the significance of that information rests on useful but imperfect behavioral associations and biological symptoms, as it seems to be in neurological lie-detection, then neuro-exceptionalism is false. While these developments can certainly raise privacy threats, those threats, as I have tried to show, are mere variations on the threats already posed by types of information we are already familiar with. This conclusion could be overturned if neuroscience identified a new kind of evidence for, e.g., lying, but that development is not on the horizon. Finally, when we consider the prospect of a type of neuro-information that identifies the content of a thought, we indeed come across a new type of information that would constitute a new kind of threat to privacy. But I have tried to argue that the prospects for developing such information are extremely remote. Outside the realm of science fiction, the case for neuro-exceptionalism is not promising.

Nevertheless, as we noted earlier, the issue of neuro-exceptionalism does not completely capture the question of whether advances in neuroscience raises new ethical problems. Even if neuro-exceptionalism is completely false, neuroscience may give us new means of acquiring nonetheless old types of information with sufficient stealth and reliability that they could constitute new and grave threats to privacy.

ACKNOWLEDGMENT

I owe thanks to David Wasserman for valuable comments on an earlier draft. Work on this article was supported in part by a grant from the National Institutes of Mental Health MH073685 and from a grant from the National Institutes of Health HG003299.

Notes