I
There is a rarely discussed openly but widely present feature of academic circles in the social sciences and humanities: the circle itself.
Insiders (researchers at universities and research institutions) have an almost instinctive aversion to outsiders: enthusiasts, independent researchers. This aversion is not always malicious. It more often manifests as a frustrated inability to communicate. Outsiders may have read widely, but their reading is scattered, eclectic, unsystematic. Insiders command a professional language and set of argumentative norms, but those norms are an illegible wall to outsiders. Both sides want to communicate, but the communication always breaks at a certain point: outsiders feel insiders are being deliberately obscure; insiders feel outsiders fundamentally misunderstand the basic concepts.
The circles are also nested at different levels. In the same discipline, the language, topics, and review standards of domestic circles and international circles are completely different. A scholar who thrives in domestic journals may not even pass initial screening in international anonymous review. Circles within circles, each with its own set of entry thresholds.
People have long accepted this state of affairs by default. Academia requires professional training, just as surgery cannot be left to amateurs. But AI's emergence has destabilized this default premise.
A clever outsider, armed with AI-assisted literature search, concept explanation, and argument structuring, can reach in a very short time a breadth of knowledge that once required years to accumulate. AI helps cross language barriers: they can read academic literature in both Chinese and English simultaneously, using AI for translation and summarization. AI fills knowledge gaps: encountering an unfamiliar concept yields instant explanation and related literature. AI can even help structure arguments: not by thinking for them, but by organizing scattered thoughts into coherent form.
In the logic of traditional academic training, this is impossible. Traditional academia is like climbing stairs: you start as an undergraduate, advancing level by level, reading the classics, learning methods, writing papers, attending defenses. Each step has its rhythm. You cannot skip. AI is like a sightseeing cable car, it can deliver an outsider directly to a mid-to-high-altitude platform. They may not be as solid as those who climbed the stairs (they missed a lot during the ascent), but they have indeed reached that height, and can see roughly the same landscape.
The question is: how should those who climbed the stairs treat those who took the cable car?
II
Academia's exclusion of outsiders is not purely arrogance. Nor is it purely a knowledge gap. The deeper mechanism is credit anchoring.
An insider's identity is, fundamentally, a credit credential. This credential is anchored by a set of social institutions: your institutional title (professor at such-and-such university), your publication record (passed peer review), your citation network (recognized by other insiders). These anchoring points collectively produce a "credibility score": what an insider says is presumed reliable until disproven. An outsider lacks this anchoring: you don't know who they are, what they've read, whether their arguments can withstand peer scrutiny. What they say might be as good in content as an insider's work, but in credit terms, it must pass an additional verification threshold.
This is why insider-outsider communication always breaks. It is not that insiders look down on outsiders. It is that the default settings of the credit system differ. When someone has no institution, no publication record, no recommenders, every sentence requires independent verification: the transaction cost of communication is too high. Most people choose not to communicate.
AI's intervention strikes precisely at the most vulnerable point of this credit system.
AI can produce texts that are formally entirely professional, yet have no accountable "author." An AI-generated literature review, with proper formatting, real citations, and fluent argumentation, is superficially indistinguishable from a scholar's work. But there is no accountable person behind it: you cannot ask "what is the basis for this judgment," you cannot trace its argumentative genealogy, you cannot invite it to a defense to answer for its claims. It is like a perfect anonymous submission, suspended forever between credible and not.
This pushes the academic credit system into a corner. The traditional system defaults to: formally professional ≈ academically trained ≈ credible. But AI breaks the equal sign between "formally professional" and "academically trained." A person with zero academic training, using AI, can produce work that is formally more polished than most graduate students' papers. The anchoring points of the credit system, institution, training, peer networks, suddenly lose their explanatory power for "formal professionalism."
III
Path one: fortify the walls.
If academia sees AI as a threat, the most natural response is to tighten the authentication requirements for "insider identity." Since AI can simulate formal professionalism, raise the bar: don't just look at how well the paper is written; require field notes, original experimental data, advisor recommendation letters. A good paper is no longer enough; you must prove you "really" did the academic labor.
This sounds reasonable. The problem is that the authentication of tacit knowledge is nearly impossible to formalize.
What is an anthropologist's "intuition" gained in the field? What is the "feel" developed in a lab through countless failures? What is the "nose" for a particular school's argumentative style acquired through years of reading? These are tacit knowledge — they cannot be written into a paper, cannot be formally verified, yet they are among the most valuable elements of academic judgment. When a committee attempts to authenticate tacit knowledge using formal indicators, the only thing it can do is substitute labor quantity for cognitive quality. Demand more pages of field notes, more experimental repetitions, longer academic CVs. This is not protecting tacit knowledge. It is symbolizing tacit knowledge: using calculable labor quantity to simulate incalculable cognitive capacity.
Carried to its extreme, this path leads to institutional isomorphism. When faced with uncertainty, academia does not actively invent new credit intermediaries; it layers additional formal standards onto existing ones. The result? AI lowers the entry threshold, but institutions raise the authentication threshold. The two cancel each other out, and those ultimately blocked are the people who lack the resources, time, money, institutional support, to meet the escalated requirements. Precisely the outsiders. Meanwhile, insiders, especially those from elite schools with research institute backing, hold a natural advantage in formalized authentication systems. AI had the potential to break down walls, but academia's defensive reaction builds them higher instead. AI does not become a liberator. It becomes an amplifier of the inequality structure.
Path two: turn toward verification.
The other path is harder, but more imaginative.
If academia does not treat AI as a threat, but as an opportunity to reconstruct the credit system, the pressure could be converted into healthy momentum. When the anchoring points of the traditional credit system, "who said it," are destabilized by AI, academia can be forced to pivot toward a more fundamental question: "how can it be verified? "
The core of this pivot is: credibility is no longer based on "you're a professor at such-and-such university," but on "in what ways can your claims be tested." Methods of verification can include open provenance tracing (what data does your judgment cite, can that data be reproduced), process transparency (what prompts did you use, what search strategies, what filtering criteria), and reproducible results (can others using the same methods arrive at approximately the same findings).
This path would not eliminate the insider-outsider distinction — tacit knowledge, disciplinary training, and peer networks do not become irrelevant simply because verification mechanisms change. But it shifts the boundary from identity authentication to process verification. An outsider, lacking institutional titles, but capable of providing a complete, traceable, reproducible research process, could have their knowledge output admitted into serious academic discussion. They aren't necessarily "right," but they can be tested, and testability is the true foundation of academic credibility.
This path requires academia to make a difficult choice: shift the anchor of credit from "who you are" to "how your knowledge can be verified." This is not merely a technical matter. It is a matter of power. The current credit system grants insiders enormous institutional dividends: your institution's prestige automatically vouches for your arguments, your predecessors have accumulated citation networks for you, your peers will not challenge your premises too harshly in your direction. Abandoning this system in favor of process verification means insiders must give up a portion of the trust they did not have to earn.
So logically, pivoting to verification is entirely possible, open-source software communities, citizen science projects, the preprint culture and open peer review practices all demonstrate that decentralized knowledge verification is not fantasy. But psychologically, building walls comes more naturally than tearing them down: defending an existing position requires no extra courage.
IV
Let's return to the cable car metaphor.
A cable car can bring tourists to a mid-altitude platform, but it cannot carry anyone into the mine. Frontier, original academic research, the kind of work that genuinely reshapes a discipline, still requires the training of climbing the stairs. You need to have read enough primary sources to recognize something new at a particular turn. You need to have repeatedly corrected your own judgment under a mentor's criticism to build an intuition for argumentative quality. You need to engage in fierce debate with peers to discover the systematic blind spots in your own thinking. These are things a cable car cannot do. AI can help you read the literature, but it cannot build your taste. AI can help you structure an argument, but it cannot make the judgment that "this direction is worth pursuing."
But the mere existence of the cable car is a healthy pressure. It lets more tourists become aware that the mine exists: outsiders have a path to roughly see what academia is discussing, where the frontier lies. This forces insiders to prove that there is genuinely something worth descending into the mine for, rather than maintaining boundaries solely through "you're not qualified to enter." A mine that can only preserve its value through closure is a mine whose value is inherently suspect.
Academic history has seen similar moments. The printing press let knowledge break free from the monastery's monopoly, forcing universities to redefine their value: no longer "we can read books others cannot," but "we can make judgments about those books' contents that others cannot." The internet let information retrieval break free from the library's monopoly, forcing scholars to redefine their value: no longer "we know where to find information," but "we know which information matters more, and is more reliable." Every technological transformation has forced academia to answer the same question: where does your value lie, when access to information is no longer the barrier?
AI is the latest, and perhaps the sharpest, iteration of this question. This time, it is not just knowledge that is openly accessible, the formal skills of knowledge production have also partially automated. Academia faces a more fundamental question: is your value "producing correctly formatted texts" (which AI can already do), or "making tested, accountable judgments within specific domains and specific communities"? If it is the latter, then walls are not your moat. Verification is.