The Era of Brain-Based Law through the Eyes of David Eagleman

10 14 Loading

Law is the distilled essence of the civilization of a people, and it reflects people’s soul more clearly than any other organism. – A.S Diamond, the Evolution of Law and Order

David Eagleman wants to bring neuroscience to a courtroom near you.

Accelerating technologies will undoubtedly challenge the basic assumptions of the mainstream culture and help shape our future values. Consequently, the legal system must adapt to reflect a new paradigm. In the following video, David Eagleman, a Baylor College of Medicine neuroscientist, discusses how our ever-expanding knowledge of the human brain may require an adjustment in our laws. If you think Dr. Eagleman is addressing these issues too soon, think again. Brain scan evidence already edged its way into an Indian murder trial, foreshadowing the use of neuroimaging in determining culpability and rational sentencing.  He also touches on perhaps the greatest concern, the ethical implications of generating consciousness de novo in machines. How will our laws deal with AI? Overall, these scenarios will force us to critically assess our value system so that society can sustain its moral footing through an era of great technological change.

In his lecture at the Royal Society of the Arts, Dr. Eagleman opens with the case of Charles Whitman (2:17), the student who went on a murder spree at the University of Texas-Austin in 1966. Following an autopsy, doctors discovered a tumor impinging upon his brain’s amygdala, prompting neurologists to consider that biologically-based dysfunction led to the tragedy. Dr. Eagleman provides additional cases of malfunctioning brains and abnormal behavior (i.e. a frontal lobe tumor causing a man to express pedophilic tendencies (4:07), Parkinson’s medication inducing compulsive gambling (5:12)). He compares these cases to automatisms, a legal concept that protects defendants if they had no control over their actions. For example, alien hand syndrome is a neurological condition in which the hand executes movements without the will of the subject. Under the automatism legal defense, if an uncontrolled hand pushes someone off a cliff, the individual isn’t held liable.

So is all criminal behavior the product of neurally-based automatisms decoupled from the will of the subject? Are we supposed to blame the person or the brain? Dr. Eagleman finds the distinction between the brain and the self to be arbitrary and based on false assumptions. In short, we are our brain. He explores this issue further at 15:48, introducing the continuum of culpability. Conditions on the far right include people with obvious damage like Phineas Gage, the perturbed man who had an iron rod lodged in his skull. Those on the far left include the “common criminal”, where the dysfunction is subtle, complex, and currently hidden from science. The line demarcates society’s present perspective on whether to blame individuals for their brain-based criminal activity, and its location is primarily determined by the available technology and our extant knowledge of brain function. According to Dr. Eagleman, this line will be pushed to the left as our understanding amasses.

The cranial continuum of culpability.

The Cranial Continuum of Culpability. A) The common criminal; B) Reggiani Martinelli claimed a brain tumor made her order a hit on her husband; C) A drug addict; D) Christ Benoit, a wrestler who killed his family in a "roid rage"; E) Phineas Gage

 

Are we exonerating every criminal that has ever lived? Well, that depends if or when we ever find the neural correlates of free will. In the mean time, Dr. Eagleman thinks that these questions are irrelevant from a legal standpoint. He adopts a utilitarian stance, positing that the only useful aim of the legal system is to decrease the frequency of criminal behavior, not to dole out blame. Starting at 18:08, he discusses the role neuroscience can play in rational sentencing. In his opinion, prison sentences should be based solely on the probability of recidivism, and he notes that risk assessments based on surveying pedophiles are already influencing prison sentences. It’s not much of a stretch to see that brain scans will be employed in the future to ascertain the likelihood of criminals becoming repeat offenders.

On this point, Dr. Eagleman and I have somewhat diverging opinions. Although, criminal law is technically intended to mediate conflicts between individual entities and the state, other parties are almost always involved. Typically, families of the murdered and abused have interests that align with the government, but this may not always be the case with strict rational sentencing. Dr. Eagleman briefly mentions the behavioral phenomenon of altruistic punishment in which individuals relinquish resources to see others disciplined. This behavioral trait could be evolutionarily viable by facilitating the psychological recovery of those affected by heinous acts. If our sentencing is to be truly utilitarian, we should consider that we may not be maximizing social benefit by focusing only on the interests of the state. If not, there could be unintended consequences, such as an increased prevalence of revenge killings (à la the Hatfield-McCoy feud).

Toward the end of the lecture (33:13), he presents a timeline for the major breakthroughs in neuro-law. The most notable of these future milestones is undoubtedly the last one, robot legislation. He poses the question: if we create a “conscious” machine and turn off the power, does that count as murder?

David Eagleman's proposed chonology of neuro-law.

Although Dr. Eagleman only briefly addresses this point, I believe the ethical and social implications of conscious machines dwarf those of brain-based sentencing. You could really teach a course on roboethics. Should machines with human or superhuman intelligence be subject to a special set of laws, or should there be an integrated legal system for humans, robots, and hybrids? In order to provide an answer based on a solid ethical foundation, we must first answer a question our species  has been dodging for quite some time:  at what point does something have the moral status of a human? We obviously have a difficult time with this one, regardless of the cultural context.  Americans have grappled with the issue of fetal rights, the Japanese have quarreled over organ transplants, and people the world over have argued for a heightened moral status of animals. If we continue to sit on this simmering problem, then it could boil over when human-like robots enter the scene. What can we do to prevent AI activists  from passing out flyers on the street?

To sort through this conundrum, I think that adopting brain-based metrics would be the best approach. One could  use the computational power of robots and human brains as a common determinant of moral status. However, there could be problems with this method because of Moravec’s paradox, which points out that rudimentary abilities require more computation than high-level reasoning. In this case, the status of “sentient being” could include far more organisms and machines than most would like. Instead, we could look at the complexity of the structure itself to determine moral status, but this could present its own issues. What if a robot has an internal system that surpasses the human neocortex in complexity but lacks the computational correlates of empathy? It seems that until we tease apart the components of the brain that make us human in the eyes of the law, determining the moral status of machines is a fool’s errand. Fortunately, according to Dr. Eagleman’s chronology, we have ample time to work this out.

Dr. Eagleman’s insights about the brain pose a serious challenge to the legal community. Neuroscience is shattering the basic assumptions of legal theory, and it will be a long time before our laws are compatible with our knowledge of brain function. It’s likely that we’ll see major court battles, and maybe even a Supreme Court case, that use fMRI images as evidence. Hopefully, the judicial system will build a comprehensive  body of precedents and our legislatures will pass rational laws that help us deal with criminal behavior using evidence-based approaches. After the dust has settled, we can prepare for the legal ramifications of advanced AI. On the other hand, I might have too much confidence in our politicians.

[Image Credits:  The Salk Institute, iStockphoto (modified), RSA Presentation]

[Video Credit:  The Royal Society of Arts (RSA)]

[Source:  The Royal Society of Arts (RSA)]

Discussion — 14 Responses

  • AdoPar29 January 30, 2011 on 6:14 am

    I recently talked with Dr. Eagleman and found out that his upcoming book, Incognito, expands on this neurolaw topic in more detail. I’ve read excerpts from it and it’s terrific.

    • Jeremy AdoPar29 January 30, 2011 on 6:16 am

      That’s great to hear. It was a real joy writing this piece and reading up on his life work thus far. Hopefully one day I’ll get a chance to meet him, too.

  • Tim Moynihan January 30, 2011 on 10:13 am

    Great article and thanks for the link to Eagleman lecture. I agree that the larger issues will emerge from AI and robotics rather than our deeper understanding of the human brain. To that end, almost all of these new technologies that could potentially become “conscious” are being born into a legal construct that is inherently distorted. They are treated as “inventions” and, as a result, subject to intellectual property law. The creators file patents, etc. to “own” the rights to these technologies. We will soon be confronted with individuals and entities standing as the owners of other conscious beings and this will present stark legal, Constitutional, and ethical concerns.

  • Pedro Galvan January 30, 2011 on 2:24 pm

    AdoPar- thanks for the rec. I just tried to find the book “Incognito” and I now understand it doesn’t publish until April. In any case, there’s some info I found here: eagleman.com/incognito. Looks like Dr. Eagleman will be speaking in many cities about this topic. Unfortunately I live in panama city so it doesn’t do me much good, but maybe you can catch him–good luck!

  • Chiwuzie Sunday February 1, 2011 on 2:18 am

    Laws are a failed attempt at solving problems. Laws do not solve problems, they are just simply: Control Mechanisms. Laws do not seek to “solve” the given problem, rather they seek to “control” human behavior — Too many egos, too many opinions, not enough fact-based solutions.

    For example, If we don’t want people to drink and drive, there are several alternative solutions we could apply, as opposed to using law enforcement and imprisonment…

    1) We could develop in-car sensors that can detect a drivers alcohol breath level. In any given case when a driver would attempt to operate a motor vehicle while drunk, our in-car sensor would immediately spot this, and pull the driver off to the side of the road to get some ‘much needed rest.’

    2) We could develop fully automated vehicles that can transport people from their drunken locations to a resting area, or to their homes safely. This could eliminate the need for anyone to even attempt to drive while drunk.

    These are just some of the possible alternative we could apply to alleviate our problems with drunk driving. The best thing about it is that our solutions can be “Tried and Tested” before fully implementing them.

    Laws do not solve problems, Technology does.

  • Jeremy February 2, 2011 on 9:59 pm

    A relevant article from Stanford’s CodeX Blog.

    https://www.stanford.edu/group/codex/cgi-bin/wordpress/?p=71

  • Anonymous February 5, 2011 on 9:36 am

    On the AI power switch issue: Unless it results in a unrecoverable crash of the underlying code, it would be more like suspended animation then death.

  • FASDMom February 8, 2011 on 7:19 pm

    Just attended a conference in Portland regarding Fetal Alcohol Spectrum Disorder and the topic was just the same. Folks born with brain damage (soaked in alcohol in utero, for example) have all sorts of “invisible” behavioral characteristics that make them rather unpleasant to be around, especially if unrealistic expectations are placed upon them. Relevant to me since I adopted a young lad 16 years ago from Russia and he clearly suffers. As do we. Once our expectations change, so too do his capabilities.