Eli Alshanetsky

As AI outputs become indistinguishable from human work, the question of whose judgment lies behind them grows more urgent. Did the student wrestle with the essay, or did the model hand it to them? Did the doctor weigh the symptoms, or did the system generate the diagnosis while they clicked through? When a polished essay no longer reveals who did the thinking, the grade above it becomes hollow, and so does the diploma. If a diagnosis can be generated by a system, the doctor’s signature risks becoming a rubber stamp. Liability remains, but the expertise and judgment that justify it erode. Even when AI’s role is small, the client can’t be sure who drove the process. Yet it isn’t just that grades and diagnoses are threatened: the credibility of our institutions is also in peril. The trust they carry cannot transfer to AI systems that bear no accountability, and vendors explicitly disclaim responsibility, warning us to use AI at our own risk. Consequently, institutional trust doesn’t migrate elsewhere. It dissolves. The answer is not to shut AI out but to govern its use by the same principle we’ve always relied on with opaque systems: institute shared rules that tie outcomes to accountable processes, even when we cannot see into each other’s minds. These are not company policies or government regulations, but rather an open protocol — like the technical standards that make email and the web possible. It’s a distributed network of trust that allows even radically different systems to interact; it’s not the only infrastructure imaginable, but it is the only kind that has ever built trust at scale. Instead of trying to halt the arms race at the level of capability development, where pressures to move ahead are overwhelming, the protocol secures the use layer. It restores trust not by stopping progress but by ensuring that progress doesn’t dissolve the institutions that sustain democracy and civic life. Education is the proving ground. Solve the problem there, and we may unlock AI’s potential where it matters most.

Deepfakes are visual and audio media that use artificial intelligence to portray people saying things they never said, doing things they never did, and experiencing events that never happened. They can be trivial (“Tom Cruise knows magic tricks?”), outlandish (“Why is Nancy Pelosi drunk on national television?”), or even dangerous (“Run, the Hollywood sign is burning!”). Because deepfakes can be both persuasive and pervasive, many commentators fear that humanity will soon take another step into the post-truth abyss. This Article evaluates the threat deepfakes pose to truth by anticipating how they will impact the area of law most directly concerned with truth: the law of evidence. Deepfakes present an obvious challenge to the administration of justice in modern courtrooms, where audiovisual evidence plays an important role. Solutions offered in past legal scholarship—like relying on experts to identify deepfakes or criminalizing deepfake production—are superficial. They optimistically assume that deepfakes will always have a tell. To truly appreciate the threat deepfakes pose, the law must brace itself for the likely prospect of what this Article calls “deepest” fakes, which will be indistinguishable in every respect from authentic media. Drawing on tools from philosophy, legal history, and technology studies, this Article demonstrates how evidence law can and likely will adapt to a world saturated with deepest fakes. Courts have long encountered the sort of philosophical skepticism that deepest fakes threaten today. In essence, the existence of deepest fakes reduces the truth value of digital media to the level of oral testimony, creative works, other easily falsifiable evidence. While deepest fakes do raise serious problems, trial procedure only needs to change if existing safeguards fail. This Article finds that deepest fakes present no different challenge for modern courts than oral testimony, paintings, and photographs presented for their early twentieth century counterparts. The safeguard then, as now, is a nuanced adversarial process that, refusing to take evidence at face value, probes each submission with contextual indicators of reliability. What emerges is an empowering picture in which human judgment, rather than blind trust in media of any sort, is the ultimate arbiter of truth.

We often get clear on our thoughts in the process of putting them into words. I investigate the nature of this process by posing the question, “Do you know which thought you are trying to articulate, before successfully articulating it?” and rejecting two answers to the dilemma it yields. The first is that the answer is yes, and that articulation is either the recollection of prior knowledge or the mere acquisition of a skill or ability rather than of propositional knowledge. The second is that the answer is no, and that your thought is unknown in that it is not yet fully realized. Clarity, according to this response, is a metaphysical property of the thought rather than the thinker’s epistemic relation to it. I offer a third solution: you start out with implicit knowledge of your thought but lack explicit knowledge of it. The process of articulation moves you from implicit to explicit knowledge.

The paper introduces a new puzzle about reflection—albeit one that is reminiscent of the famous paradox about inquiry in Plato’s Meno. We often make our thoughts clear to ourselves in the process of putting them into words. Our puzzle is that, on the one hand, coming to know what we are thinking seems to require finding words that would express our thought; yet, on the other hand, finding the words seems to require already knowing what we are thinking. I argue that the puzzle cannot be solved by accounting for the knowledge that we gain in such cases on the models of self-interpretation and self-constitution. The primary purpose of introducing the puzzle is to provide a tool for systematically investigating this category of non-interpretive and non-constitutive self-knowledge, and I conclude with some words about the scope of the category and the value of finding a solution.

Articulating a thought can be astoundingly easy. We generally have no trouble expressing complex ideas that we have never considered before, though not always. Articulating a thought can also be extremely hard. Our difficulties in articulating thoughts pervade many aspects of philosophical inquiry, as well as many ordinary situations. While we may overcome some of the challenges through education and practice, we cannot do away with them altogether. And the hardest thoughts to articulate often come to us unbidden: as we neither assemble them from other thoughts nor get them from any source of external information. They can come from us freely and spontaneously, and frequently we articulate them in order to find out what they are. In many cases, we would not bother articulating our thoughts if we already had this knowledge--yet, when we find the right words, we can often instantly tell that they express our thought. How do we manage to recognize the formulations of our thoughts, in the absence of prior knowledge of what we are thinking? And why is it that producing a public language formulation contributes in any way to the deeply private undertaking of coming to know our own thoughts? In Articulating a Thought, Eli Alshanetsky considers how we make our thoughts clear to ourselves in the process of putting them into words and examines the paradox of those difficult cases where we do not already know what we are struggling to articulate.