THE STUDY GROUP
Chapter Two: The Calibration Problem
The first meeting with James Harrow was Thursday at ten, in his office on the fourth floor of the Linden Building, and Nadia arrived two minutes early to find him already there, already absorbed in something at his desk.
"Sit," he said, without looking up.
She sat.
He finished reading, whatever it was, with the complete attention of a man who did not half do things, then set it aside and looked at her.
"Tell me about the mismatch signal question," he said. "Not the version from your research statement. The actual version. What you think when you're not writing for an admissions committee."
She looked at him.
"The actual version is messier," she said.
"Good," he said. "Messy means real."
All right, she thought. "The actual version is that I don't know whether social prediction error is computationally distinct from sensory prediction error, or whether I just think it is because the social domain feels qualitatively different to experience. Which means I'm not sure if I'm studying a real distinction or a phenomenological intuition that may not map onto underlying architecture."
James looked at her steadily.
"That's better," he said.
"It's not a very confident position."
"It's an honest one," he said. "The question is whether you can convert the honest position into a research programme." He leaned forward slightly. "What would it take to convince you the distinction is real?"
"A dissociation," she said immediately. "Conditions where sensory prediction error is intact and social prediction error is impaired, or vice versa. Patient populations or experimental manipulations that pull them apart."
"And if you find no dissociation?"
"Then either they're the same computation, or my paradigms weren't sensitive enough to detect the difference," she said. "Which means I accept the null hypothesis or redesign."
"You're comfortable with null results." Not a question.
"I'm not comfortable with them," she said. "But I think null results are real results, and I'd rather have a real null than a false positive."
He was quiet a moment, the specific quality of a man reassessing.
"Most students at this stage are afraid of null results," he said.
"Most students at this stage haven't spent four years working up to one question," she said. "I know what I'm asking. If the answer is no, I want to know it."
Something shifted in his face, the controlled assessment giving way to something warmer. Not warm the way Finn was warm, nothing so abundant, but real.
"We're going to argue about the methodology," he said.
"I expected that."
"Your proposed comparison paradigm has a confound," he said. "The social stimuli and the sensory stimuli aren't matched on complexity."
"I know. That's the problem with social stimuli, they're inherently more complex. The question is whether you control for complexity or accept it's constitutive of the phenomenon."
He looked at her for a long beat.
"You've been sitting with this objection."
"I wrote the paradigm," she said. "I've sat with all the objections."
"What are the others?"
She told him. Seven objections she'd identified, with the responses she'd built for each. He listened without interrupting, which was its own form of intensity, the attention of a man not waiting for a gap in which to speak but actually absorbing what she said. When she finished he was quiet for perhaps thirty full seconds.
"Objection four," he said. "The one about baseline neural activity in social versus non social contexts. Your response is adequate but not sufficient."
"I know. I don't have a sufficient response to objection four yet."
"What are you going to do about it?"
"Find one," she said. "Or restructure around it."
He looked at her a moment longer, and then, for the second time, the sudden full smile that remade his face.
"Right," he said. "Same time next week. Bring me a draft of the literature review, just the first twenty pages. I want to see how you handle the Friston to Clark transition."
"The Friston to Clark transition is contested."
"I know," he said. "I want to see how you handle contested terrain."
She gathered her notebook and stood.
"Nadia," he said, as she reached the door.
She turned.
"The honest position," he said. "Keep it. The field is full of very confident people who are wrong. Intellectual honesty is rarer, and more useful."
She looked at him.
"Thank you," she said, and meant it more than the two words could carry.
She walked down the Linden Building stairs thinking about calibration, the slow process by which a prior was dragged toward accuracy, belief brought into correspondence with evidence. James Harrow, she decided, was a calibration machine. Everything about him pointed at the same question: is this true?
* * *
Finn's office was warmer than James's, in every sense.
Not physically, the Linden Building had its own highly opinionated radiators, but the space itself. Bookshelves with the organic disorder of someone who actually read them rather than curated them. A small plant on the windowsill doing its best. Student artwork on the wall, pieces she recognised as the kind of thing PhD students brought their advisors as thank you gifts at the end of a dissertation, evidence of people who had been loved here before her.
"You've survived James," Finn said when she appeared in his doorway.
"He wasn't as frightening as his reputation."
"He's exactly as frightening as his reputation," Finn said. "You're just not frightened by the right things yet." He gestured at a chair. "Sit. How do you take your coffee?"
"White, one sugar. You don't have to."
"I make coffee at two," he said. "It's what I do at two. Tell me about your undergraduate research."
She told him, while he made coffee with the easy familiarity of a man who had done it a thousand times. He listened differently from James: not the intense receiving of data, but something conversational, following the thread and picking it up at unexpected points, asking less is this right and more what does this mean to you.
"The clinical angle," he said, handing her the cup, his fingers brushing hers a half second longer than the handoff required. "Tell me about that part."
She noticed the warmth of his hand and set it aside, the way she set aside things that were not useful.
"My cousin was diagnosed with autism at seven," she said. "I was twelve. I watched the way she processed social situations and thought, there's a different logic here. Not a deficit. A different inference process." She paused. "The deficit framing bothered me even before I had the vocabulary to say why."
"And now you have the vocabulary," Finn said.
"The predictive processing framework is the vocabulary. If social cognition is inference, if we're all just running different models, then the question isn't which model is correct but how different models interact. Where they produce mismatches."
"Nadia."
"Yes?"
"That's the most interesting research question in the department," he said.
"James would argue."
"James argues with everything," Finn said. "That's his job. His job is to take your question and make sure it's strong enough to hold its own weight." He looked at her steadily. "Mine is different."
"What's yours?"
"To make sure you remember why you asked the question in the first place."
She held her coffee cup.
"What's Dr. North's?" she said, before she could stop herself.
Finn looked at her with a specific quality, not alarm, not suspicion, but careful attention.
"Elias makes sure the mathematics are doing what you think they're doing," he said. "The formal structure. If you claim a Bayesian model, he'll make sure it's actually Bayesian and not just borrowing the word." He paused. "He's also the one who'll push you on what's novel. James will ask if it's true. I'll ask if it matters. Elias will ask if it's been done before and what you're adding."
"That sounds useful."
"It is. The three of us are somewhat overdetermined for this kind of research. We don't agree on everything. That's the point of the tri advisor structure, three different sets of pressure, and you have to find the approach that holds up under all three."
"That sounds like a lot of pressure."
"It is," he said. "The students who do well in this structure are the ones who can hold their position against reasonable opposition." He looked at her. "I don't think you'll have difficulty with that."
"You don't know me yet."
"No," he agreed. "But I read your research statement four times, and you pushed back against the two reviewers' objections in the supplementary materials in a way that was direct without being defensive." He smiled. "That's a specific skill. Some people learn it. Some people have it."
She thought about James and the objections, the way she'd laid them out already expecting his challenge, the warmth of Finn's hand still a faint ghost on her fingers.
"The three of you," she said. "You work well together?"
"We work interestingly together," Finn said, which was a different answer. "James and I have been colleagues for seven years. Elias joined five years ago." He paused. "The three of us think differently enough that it's productive."
"But?"
He looked at her steadily.
"There's no but," he said. "We're good colleagues. We occasionally disagree about methodology and university politics, but none of that is relevant to your degree."
She remembered what he'd said on orientation day. Don't let us argue you out of it. The us had included all three of them.
"Finn," she said.
He raised an eyebrow at the first name.
"You said us on orientation day. You said us."
He looked at her for a moment.
"I did," he said.
"Is there something I should know about the three advisor dynamic?"
"You'll see it for yourself," he said. "That's probably more useful than me telling you." He picked up his coffee. "The thing to know is that all three of us want you to succeed. The route we'd each take to get you there looks different. That's not a problem. That's the structure working."
She sat with that a moment.
"And if I find myself between two contradictory pieces of advice?"
"You'll find that happening about once a week," he said. "In those cases, come to me. I'll help you find the synthesis."
"Not to James?"
"James will tell you which position is correct. That's useful when one of them clearly is. For the genuinely ambiguous cases, I'm a better sounding board."
"And Dr. North?"
Again, that specific careful attention.
"Elias will tell you what the models say," he said. "The formal logic. Useful for a different set of problems."
She nodded.
"Same time next week?" he said.
"Same time."
She walked back to her temporary housing under the long shadows of a September afternoon, thinking about three advisors and three different kinds of pressure, and the particular problem of having already, without intending to, established something with one of them that was not quite a professional relationship, and that she could still feel in her chest like a held breath.
The prior was wrong, she thought again. The evidence is updating the model.
The question is what the updated model contains.
* * *
Elias North's office hours were Tuesdays, two to four.
She had not gone the first Tuesday. She had told herself this was because she didn't have anything specific to discuss yet, which was true, technically, but the more honest version was that she wasn't sure what was going to happen when she walked into his office, and she found uncertainty about social prediction difficult to sit with, given that it was her dissertation topic, and harder still to sit with when the uncertainty had a face she could not stop remembering.
The second Tuesday she went.
His office was different from the others the way he was different from the others, not disordered like Finn's or controlled like James's, but spare, the specific spareness of a man who had thought about what he needed and removed everything else. A whiteboard covered in equations that were not, as far as she could tell, from anything currently in the literature. A desk cleared except for whatever he was currently working on.
He was at the whiteboard when she knocked on the open door.
"Nadia," he said. He didn't sound surprised. He also didn't turn immediately, finishing his notation first, and then turning.
"Dr. North."
"Elias," he said. "You've been calling the others by their first names since the second week."
"You haven't offered it."
A pause.
"Elias," he said again, and something in the way he repeated his own name, slower, lower, landed somewhere low in her stomach.
She came in and sat, because there was a chair and sitting felt less exposed than standing in front of him.
"How's it going?" He had moved away from the whiteboard but hadn't sat, leaning against the edge of his desk, closer than a desk should put you, further than a conversation usually sat.
"Well. James and I are in productive disagreement about the methodology."
"That's his normal mode," Elias said. "How are you finding Finn?"
"He's different. Easier, in some ways."
"Yes," Elias said. "That's accurate." His eyes stayed on her, unhurried. "You've been avoiding this office."
She held his gaze, refusing to look away first.
"I've been settling into the programme."
"Yes," he said. "That's also true." He folded his arms. "Nadia. The situation on the plane is unusual. I want to be clear it doesn't affect the professional relationship."
"I know."
"You're one of three students in the programme I'm responsible for," he said. "The work comes first."
"Of course."
"Good." He pulled up a chair and sat, and suddenly they were at the same level, his knees a careful, deliberate distance from hers, close enough that she was aware of the exact measurement of the gap. "Then let's talk about your Bayesian framework. The inference model you proposed makes a specific assumption about the precision weighting of social signals. Walk me through it."
She walked him through it. He listened and asked precise, occasionally uncomfortable questions, the kind that revealed an assumption she'd made without examining it. Twice he went to the whiteboard and wrote something, and once, reaching past her to reclaim a marker from the tray beside her chair, his forearm brushed hers, and neither of them acknowledged it, and both of them, she was almost certain, noticed.
"The assumption," he said, at the end, "is defensible, but you'll need to justify it explicitly. If you leave it implicit, a reviewer will find it."
"I know. That's why I'm here. I wanted to work through it with someone who would find it."
He looked at her.
"You came to office hours specifically to have me find the problems in your framework?"
"That's what office hours are for."
"Most students come to office hours to have their existing ideas validated."
"That doesn't seem useful."
He was quiet a moment.
"No," he said. "It isn't." Another pause, longer this time, his eyes not leaving her face. "You were doing this on the plane. You do it naturally."
"Do what?"
"Seek the challenge rather than the confirmation," he said. "It's an unusual trait."
She thought about the research statement and objection four and the thing she hadn't resolved, and the faint, persistent awareness of his forearm, still warm where it had touched hers.
"My research is literally about how minds update their models," she said. "It would be somewhat inconsistent to refuse to update mine."
The almost smile, the first she'd seen from him since the plane, since it had become a complicated situation rather than a pleasant conversation, and it did something unfair to the lower half of her stomach.
"Fair point," he said.
They sat with the whiteboard between them and the September afternoon going dark outside, and she thought: this is what it will be. The work, in this room, with this particular quality of attention that is both easier and harder than the others. Easier because I recognise it. Harder for the same reason.
"Same time next week?" she said.
"Same time," he said. "And Nadia."
She looked at him.
"The framework has problems," he said. "They're solvable problems. The framework itself is sound."
She stood.
"That's the most direct positive feedback I've heard from you."
"I'm sparing with it," he said. "That's a feature, not a fault."
She walked to the door, and at the threshold something in her, some reckless, honest impulse she usually kept leashed, made her stop.
"Elias."
He looked up.
"I'm glad it's you," she said. "The person from the plane. Not a stranger, but you."
He went very still, the kind of stillness she would come to recognise as the opposite of indifference.
"That's a complicated thing to say."
"I know. I'm saying it anyway. You told me on orientation day the situation was going to be complicated. I'm acknowledging that it is, and also that I'm all right with it."
He held her gaze for a long moment, long enough that she felt her pulse pick up in a way that had nothing to do with academic anxiety at all.
"The work comes first," he said, but his voice had dropped half a register, and it did not sound like a refusal so much as a man reminding himself of a rule he was working hard to keep.
"I know," she said. "That's also all right."
She went, her heart loud in her own ears the whole way down the corridor.
The prior, she thought, walking. The prior says: this is your advisor. The prior says: this is professional. The prior says the complications outweigh the.
But priors update, she thought. That's what they do.
I'm studying the mismatch signal. The moment when the world fails to match the model.
I think I've been conducting fieldwork.
* * *