The next step.
Use AI to understand a complex paper — then design the experiment that comes next.
Use AI to truly understand a complex, recent paper — then do what a good scientist does next: design the experiment that should come after it.
Why it matters
Papers keep getting longer and more technique-dense, and no one is fluent in every method. Being able to rapidly metabolize an unfamiliar paper — and then push past it to “so what would I do next?” — is the core engine of research: journal clubs, grant aims, rotation projects, collaborations. This is also a clean test of AI for high-stakes thinking, where a confident-but-wrong explanation of a method, or a “novel” experiment that was actually published years ago, has real costs.
The paper
You’ll work from a recent paper on the circuit basis of placebo analgesia:
Livrizzi G, Chang-Weinberg J, Johnson DA, et al. Top-down control of the descending pain modulatory system drives multimodal placebo analgesia. Neuron (2026). doi:10.1016/j.neuron.2026.03.025
The download is the open-access preprint of the same study (its title omits “multimodal”); cite the published Neuron version above. Paste the PDF or its text into your AI tool to begin.
Run of show
- 0:00–0:05 · Challenge introduction (5 min)
- 0:05–0:20 · Work in your group (15 min)
- 0:20–0:22 · Post your best prompt (2 min)
- 0:22–0:32 · Share & debrief (10 min)
- 0:32–0:35 · Reset (3 min)
The challenge
This challenge has two parts.
Part 1 — Understand it. Use AI to do a rapid analysis of the paper and create your own summary of the study and its findings — in a format that lets you explain the experiments that were done and their major findings, not just restate the abstract. Think about what would actually be useful to you or your lab: a technique glossary? An experiment-by-experiment logic map? A one-slide model figure in words?
Part 2 — Extend it. Come up with one follow-up experiment that tests an assumption or finding in their work, or extends it in a meaningful way.
As you go, push the AI past surface answers:
- What kind of summary would be genuinely useful beyond the existing abstract?
- What limitations and caveats can AI help you uncover that you wouldn’t otherwise notice?
- Can AI help you check whether your proposed experiment is new — or already published by someone else?
- Can AI help you evaluate the feasibility and logic of your proposed experiment?
Bad prompt to better prompt
Why it disappoints: you get a polished paragraph that mirrors the abstract, glosses over the techniques, never explains why each experiment was run or what it rules out, and surfaces none of the assumptions you’d need in order to design what comes next.
Follow up with: “Propose two follow-up experiments that each test one of the assumptions in section 3. For each: hypothesis, design, predicted result, the main confound, and how I’d check whether something like it has already been published.” Then verify any cited prior work yourself before trusting it.
Debrief questions
- What summary format did your team build — and could you now explain the paper’s key experiment without re-reading the AI’s output?
- Where did the AI misdescribe a technique or overstate a finding? How did you catch it?
- What assumption or limitation did your follow-up experiment target — and why is it the interesting one?
- Did AI help you check whether your experiment was novel? Did it find prior work, or just claim there wasn’t any?
- How feasible is your proposed experiment, really? What did the AI miss about cost, time, or confounds?