Neuroblame?

Written by Stephen Rainey

Brain-machine interfaces (BMIs), or brain-computer interfaces (BCIs), are technologies controlled directly by the brain. They are increasingly well known in terms of therapeutic contexts. We have probably all seen the remarkable advances in prosthetic limbs that can be controlled directly by the brain. Brain-controlled legs, arms, and hands allow natural-like mobility to be restored where limbs had been lost. Neuroprosthetic devices connected directly to the brain allow communication to be restored in cases where linguistic ability is impaired or missing.

It is often said that such devices are controlled ‘by thoughts’. This isn’t strictly true, as it is the brain that the devices read, not the mind. In a sense, unnatural patterns of neural activity must be realised to trigger and control devices. Producing the patterns is a learned behaviour – the brain is put to use by the device owner in order to operate it. This distinction between thought-reading and brain-reading might have important consequences for some conceivable scenarios. To think these through, we’ll indulge in a little bit of ‘science fiction prototyping’.

Science fiction prototyping is a way to project reasonable expectations for technologies into a near-future so that probable consequences from those technologies can be anticipated and addressed in the present, if necessary. In terms of BMIs and BCIs, we can imagine easily that, in a near-future scenario, they will migrate from therapeutic settings into the more everyday. This is already the case, to some extent, with some technologies. But we can imagine this being extended further.

Rather than bio-mimetic prostheses, replacement limbs and so on, we can predict that technologies superior to the human body will be developed. Controlled by the brains of users, these enhancements will amount to extensions of the human body, and allow greater projection of human will and intentions in the world. We might imagine a cohort of brain controlled robots carrying out mundane tasks around the home, or buying groceries and so forth, all while the user gets on with something altogether more edifying (or does nothing at all but trigger and control their bots). Maybe a highly skilled, and well-practised, user could control legions of such bots, each carrying out separate tasks.

Before getting too carried away with this line of thought, it’s probably worth getting to the point. The issue worth looking at concerns what happens when things go wrong. It’s one thing to imagine someone sending out a neuro-controlled assassin-bot to kill a rival. Regardless of the unusual route taken, this would be a pretty simple case of causing harm. It would be akin to someone simply assassinating their rival with their own hands. However, it’s another thing to consider how sloppily framing the goal for a bot, such that it ends up causing harm, ought to be parsed.

Let’s imagine I send my neuro-controlled grocery-bot out to get food for the evening. Through insufficient foresight, I fail to control the bot well enough and it mows down 40 people at the deli counter as it pursues my goals. If we think about this in terms of legal culpability, I as the neuro-controller, don’t have the mens rea or actus rea as I neither intend to harm the people, nor do I physically carry out the harm. Yet the cause of the harm was indeed under my control. My neural activity was, or at least ought to have been, steering the bot’s activity. My neural patterns were insufficient to avoid the harm occurring.

The user of a neuro-controlled bot is blameworthy for the production of non-optimally instrumental neural patterns of activity. This is an unusual way to think of blaming people, but it seems appropriate in cases of BMIs gone awry. It might be the case that BMIs, and such technologies, ought to be thought of in similar ways to driving. If I am inattentive at the wheel, I’ll get points on my licence, I might be required to attend extra courses on road safety, I may even have my licence revoked and serve time in prison. In the sci-fi scenario, the sloppy brain-work of the bot-controller leads to harm. The user is due some negative consequences for that.

Back in the present, it might be useful to think about BMIs etc. very explicitly in terms of instruments driven by the use of the brain. In terms of driving, if I physically move my hands such that they steer the car in a way that violates the rules of the road, or harms somebody, that physical act leads to wrong. This is what ought to be transplanted into neuro-control cases – the physical manifestation of neural activity, in being learned, and practiced, is an act just like steering the car with my hands. We can’t easily look at the neural case, as it’s under the skull, but it is nonetheless a physical act.

While it probably does seem unusual to think of someone as using their brain in this instrumental sense, it will likely serve us better in conceiving of these kinds of technologies. Talk of ‘thought controlled’ devices, on the other hand, misleads. If we talk in that way, it seems harder to make sense of blaming me for the harm my bot causes at the deli. My thought was ‘get some prosciutto’, but the act, the blameworthy part, was a realisation of neural activity that didn’t provide enough robot control to avoid the other customers.