Whose name owns the decision?
A specific claim is now shaping this argument: a commentator said that Opus 4.8 outperformed human experts as peer reviewers on 124 Harvard social science papers. The same description says those papers spanned statistical analyses, formal models, and qualitative work. That matters because it places the claim inside scholarship as it is actually practiced, across different methods, not in a narrow demo built around one style of paper.
The risk here is easy to misstate. The main risk is not that machines are entering scholarship. Machines have already entered drafting, checking, summarizing, and screening in many research workflows. The sharper problem begins when journals adopt machine judgment faster than they rewrite who is answerable for accepting, rejecting, and defending a paper.
A review system is not only a quality filter. It is also a chain of responsibility. A recommendation moves upward, then a person signs. That signature tells authors, editors, readers, and institutions where the judgment finally lives. If the recommendation is increasingly produced by a system while the formal signoff stays unchanged, the visible owner of the decision can drift away from the real source of judgment.
That drift is where accountability gets thin. A journal can say a human editor made the call. But if the editor is mostly ratifying a machine ranking, machine summary, or machine recommendation, then the practical decision may already have moved. The issue is not whether the machine is always right or always wrong. The issue is whether the institution has said, plainly, who must defend the result when a paper is accepted, rejected, appealed, or later challenged.
One response to the review claim made the pressure explicit: the same commentator wrote that academics need to adapt the practice of science to AI quickly. Taken as reasoning rather than proof, that response points in the right direction. Capability changes governance. If review quality changes, institutions have to decide whether the human signature still matches the real judgment.
Match signature to real judgment
The prepared move is not to argue in the abstract about whether AI belongs in peer review. The useful move is to locate the exact point where recommendation and signoff may already be separating. In one journal, that may be the editor who receives an AI-generated assessment before reading the full paper. In another, it may be the reviewer who starts from an AI draft and then lightly edits it before submission. In both cases, the question is the same: who is actually accountable for the final decision as it is really being made?
A workable answer starts small and concrete. Name the stages in the review path. Note where AI produces a ranking, summary, critique, or draft. Then identify which human role is expected to stand behind the outcome in public and in dispute. If that person cannot explain the reasoning without leaning on the system's output, the process may need redesign rather than a new disclaimer.
That redesign does not require banning AI or pretending nothing changed. It means making the official structure match the real one. A journal might decide that editors can use AI support but must independently state the decisive reasons for acceptance or rejection. It might decide that AI-generated reviewer drafts must be disclosed inside the editorial process. It might create a rule that the signatory is responsible only if they can articulate the grounds of the decision in their own words. These are governance choices. They are still human choices.
This week, identify one review process where the recommendation and the signoff may already be drifting apart, and write down who is actually accountable for the final decision before the mismatch goes quiet.
Human Runtime Formula™ is designed for the mental edge you need to survive in an AI-run world.