by Tim Sommers
In 2013, a team of researchers at MIT successfully implanted a false memory in mice of being shocked and then “reactivated” that implanted memory. As a result, the mice responded to fear conditioning without actually being shocked.
In 2018, working with epilepsy patients, a team of researchers at USC recorded the firing patterns as memories were encoded, built a model of how the hippocampus transforms input patterns into output patterns, and then stimulated the model to predict “correct” patterns. Reportedly, they improved memory performance in patients by 34–37%.
Of course, real-life trials are well behind fiction. In William Gibson’s seminal cyberpunk novel Neuromancer small chips called “microsofts” can be slotted into a socket behind your ear and give you instant fluency in a language or a technical skill. Going all the way back to 1932, Aldous Huxley’s Brave New World, featured “sleep-teaching drills” that program citizens to have the correct beliefs about the state. It’s conditioning through the ears while unconscious, not writing to the brain directly, but still…
Will we ever be able to download knowledge directly into a human brain the way we download software into a computer? At this point, this seems not only conceptually possible, but empirically likely.
But I am a philosopher, not a neuroscientist, and I want to look at the epistemological questions this possibility raises, rather than its feasibility. Despite its science-fictionyness, there’s a whole line of thought experiments related to the question of whether or not implanted or installed knowledge can really be knowledge. Let’s go there. Be warned, however, we are going deeper into the weeds than we normally do.
Let’s start with Alvin Plantinga’s brain lesion. I mean, brain lesion example. What if there were a type of brain lesion that, among other things, caused its victims to believe that they have a brain lesion? Such a lesion, let’s imagine, always results in the victim believing they have a brain lesion. Yet, almost no philosopher thinks that the victims in these cases know that they have a brain lesion. If knowledge is, as most epistemologists think, justified true belief, we could describe what goes wrong here as the true belief that they have a brain lesion is not justified, or justified in the right way, to count as knowledge. The brain lesion is damage, a source of malfunction. The weird coincidence that the lesion makes you think you have a brain lesion is more akin to a hallucination than knowledge. Reliabilists are epistemologists who say that what justifies true beliefs is the operation of some reliable process. So, in the lesion case, the true belief is not knowledge because it results from pathology rather than a properly functioning process.
But imagine undergoing a procedure marketed as “direct learning.” A device maps the neural realization of belief content and installs new, well-established, and coherent beliefs directly into your brain. Expert teams curate the provider’s catalog and update it continuously. Its error rate is lower than that of any university course. You might select the module “Biochemistry I” and wake an hour later with several thousand new, true propositions about biochemistry available to you. The beliefs are stable, mutually integrated, and integrated with your prior beliefs. You can deploy them fluently in reasoning and action.
On the other hand, you have seen no evidence for these beliefs, have performed no experiments, have read no textbook, and heard no lecture or testimony of any sort. If someone asks you how you know that the citric acid cycle occurs in the mitochondrial matrix, about all you can say is, “I did the module.” The question is, if the process is sufficiently reliable, isn’t it, in fact, a new and somewhat exotic, but quite reasonable, way of gaining knowledge?
Consider “BonJour’s case” of a clairvoyant named Norman. His clairvoyance is perfectly reliable, but he has no evidence that he possesses it and no other grounds for the beliefs that result from it. And yet he does believe them. And they are all true. This seems like a clear case where the beliefs he forms clairvoyantly are clearly not like knowledge.
Even if that is true, the installation case is not like the BonJour case. It’s not based on a mysterious power whose reliability is opaque, not just to him, but even to readers. The installation process’s reliability is publicly documented and, indeed, known to its beneficiaries.
Consider instead Lehrer’s Truetemp, a character who has had (without his knowledge) a pin-sized device installed in his scalp that transmits temperature readings to his brain and makes him think about the temperature constantly. Does he always know the temperature?
Since he did not undergo the implantation voluntarily, is unaware of the process, and has not endorsed it, he is, at a minimum, alienated from this process of belief formation. We might think that he doesn’t really know the temperature, even if he can report it accurately.
But people who undergo the installation process for knowledge are not alienated from belief formation the way Truetemp is. In fact, they signed up for it.
Finally, Gettier-style cases raise issues. Gettier attacked the idea that knowledge is justified true belief, raising a whole new type of problem with it. Suppose, for example, you had a light blue Cali Classic scooter parked outside your back door, but unbeknownst to you, it had been stolen. But also, unbeknownst to you, you won a contest, and the sponsors dropped off your prize, a new light blue Cali Classic scooter, out back. If you tell someone, “Yeah, I have a light blue scooter around back,” this is true, you believe it, and here’s where it gets weird: it’s justified. Sort of. Another has replaced one scooter fitting that description, but you seem, belief-wise, to be tracking the wrong thing, justification-wise. And yet you do have a justified true belief that your scooter is out there waiting for you.
It doesn’t matter, though. Unlike Gettier cases, nothing about the etiology of the beliefs of people who have had them implanted involves luck, accident, or inference through falsehood.
The point of all these examples, you see, is to strip away the usual ways of dismissing implanted beliefs. Our hypothetical subject’s reasoning about her beliefs is not hostage to luck or accident. She knows the source of the belief and has reason to believe it is reliable. If there is something epistemically deficient about her knowledge state, it lies elsewhere than in these traditional hypotheticals. In fact, I will argue that it lies in the relation between the subject and her beliefs, and not in the beliefs’ content or structure or genealogy.
To be clear, it seems clear that implanted beliefs are not necessarily irrational or unfit to act upon. Indeed, I don’t deny that we might be better off having them and using them. But I still want to know if we know the beliefs we have implanted. Next time, I’ll propose what I think is a novel approach to saying why implanted beliefs don’t initially constitute, but can become, knowledge. If that succeeds, then here’s the real payoff: a new condition on knowledge that goes beyond justified true belief.