Answering Questions about the Academy

Alex sent me some questions about dishonesty in the academy. I assume that they were inspired by the essay Why Most Academic Reasearch is Fake. In this post, I will answer his questions.

What ‘big important problems’ are ignored by academic research?

There are many examples of academic fields ignoring core issues so that they can focus on minutia.

The notion of utility, which is widely used in philosophy and economics, is almost entirely taken for granted. Economists assume that utility can be measured with monetary metrics, such as GDP. Important questions, such as “What is utility?” and “Can we quantify utility?” are ignored, despite the notion of utility being essential to both moral philosophy and economics.

Evolutionary theory is not well-defined in academic biology, despite being the core theory. Meanwhile, evolutionary biologists spend their time contriving Rube Goldberg mechanisms to avoid the implications of evolutionary theory, rather than trying to understand it. A huge amount of bullshit research has been generated in the field, because the core theory is either ignored, denied or misunderstood.

The problem of meaning has been ignored or side-stepped by psychology, cognitive science and philosophy to a large extent. The correspondence theory of truth is essentially an attempt to dodge the problem of meaning.

In cognitive science and AI, there was a huge amount of research related to logical-symbolic reasoning, which never led to any applications, because logical reasoning isn’t very useful by itself. Most AI research had nothing to do with the development of LLMs, or anything else that people use. AI researchers plowed ahead solving little made-up problems, while leaving big questions unanswered, such as “What is intelligence?”, “What is knowledge?”, “How do we acquire knowledge from experience?” and “How do ideas represent reality?”.

In academic philosophy, there is a discourse based on the definition of knowledge as justified true belief. But the concepts of justification, truth and belief all presuppose a theory of knowlege, so this definition is a vacuous circle, and so is all discourse based on it. For example, see the Gettier problem.

What is the purpose of life? What is value? These are important questions about the human condition, but they tend to be ignored within the academy. Instead, academic philosophers and psychologists presuppose some notion of flourishing, happiness or well-being, which is typically left undefined. Steven Pinker and Sam Harris do this, for example.

Academic moral philosophy largely consists of moral apologetics: a post hoc rationalization of popular moral intuitions. The majority of academic philosophers identify themselves as moral realists. (See 2020 PhilPapers Survey.) This shows the irrationality and stagnation of academic philosophy.

The question raises bigger questions. Why do academics tend to avoid big problems? Why is the academy stagnant?

Across fields, most academics avoid core theoretical questions. It is easy to write and publish a paper on some small incremental “brick” of knowledge that fits into an existing theory. It is hard to write and publish a paper that contains substantive original research.

Many papers are based on already published papers. The old paper provides a template for the new paper. The new paper extends the research in the old paper, or tweaks it in a minor way. The citations can be copied from the old paper, often without being read. This is not overt plagiarism, but it is copying the work of others. However, it is not frowned upon. The new paper will cite the old paper as an inspiration. The new paper will probably be reviewed by the same people who reviewed the old paper, so it is likely to pass peer review. It might even be reviewed by the authors of the old paper, who will be pleased that their work is being cited.

It is much harder to publish research that is substantively original, especially if it challenges an existing theoretical framework.

We can divide research into two types: theoretical and detailed. Theories provide the foundation and framework for detailed research. Detailed research is incremental, while theoretical research has the potential to be revolutionary.

Metaphorically, we can think of detailed knowledge as a pile of bricks that sits on a theoretical foundation. Each brick is held up by the bricks below it and by the foundation. The more bricks are stacked on top of a brick, the harder it is to move or replace that brick. As the bricks pile up, it becomes harder to change the foundation.

As detailed research accumulates on top of a theory, it creates theoretical inertia. The theory becomes a dogma that is held in place by vested interests.

Some detailed knowledge is theory-independent, so it does not cause theoretical inertia, and it can survive a theoretical revolution. For example, astronomy requires the careful observation of objects in space. Let’s call this type of scientific research “data collecting”. Data collecting is very different from theoretical science. We should not expect data collectors to make good theorists, or vice versa.

Most detailed knowledge, however, depends on theories and creates theoretical inertia.

Academic fields develop in stages that are similar to the stages of colonization. First come the explorers: men who are attracted to the unknown. They create the initial theories. Then come the pioneers. They fill in the landscape. They bridge the gap between theoretical research and detailed research. They are not dogmatic about theories, but they are not revolutionary either. Then come the settlers, who are only concerned with detailed research. They want to farm some little portion of the theoretical landscape. Finally, the opportunists arrive. They have no interest in the details or the theory. They just want to leech off the system.

These stages correspond roughly to generations, although the final state of a field is usually a mix of settlers and opportunists. But if there is little or no detailed research to do, a field can become almost entirely populated by opportunists.

Once a field has become settled, it tends to become stagnant, due to theoretical inertia. Many people are invested in the theoretical framework, so there is strong resistance to changing it. The field degenerates into a circular discourse game or the accumulation of minutia, or both.

Theoretical revolutions are necessary to prevent stagnation. But it is not easy to overturn an existing theoretical framework.

In The Structure of Scientific Revolutions, Thomas Kuhn pointed out that scientific revolutions involve more than just rational persuasion. A scientific revolution has to overcome the inertia created by academic investment in the existing theory. As an institution, science is not entirely rational, because rationality is not the only basis for selecting scientific theories.

When a scientific revolution occurs, it sweeps away the detailed knowledge of the previous theoretical framework.

In the geocentric theory of the universe, epicycles were an important type of detailed knowledge. When that framework was replaced by the heliocentric theory, the epicycles were discarded.

For many years, Freudian psychology was a major theoretical framework within academic psychology. Thousands of research papers were published within that framework. Today, Freudian psychology is no longer fashionable, and that research sits on shelves in university libraries, gathering dust.

How can the state create incentives for the generation of real academic research?

First, the state should reduce academic funding. A reduction in the size of the academy (or at least the research side of the academy) would make it more productive.

To some extent, the academy is a rent-seeking cartel. The state should break up this cartel, and ensure that there is a fair competition between ideas. This would help to prevent stagnation.

Knowledge production is analogous to economic production. Communism doesn’t work for either. The state can play a role in defining the rules of the game and enforcing those rules.

To generate intellectual progress, we need a process with three components:

• Innovation: New ideas are generated by creative individuals.
• Communication: Ideas propagate. New ideas can compete with existing ideas.
• Selection: Ideas are selected based on norms of rationality.

This is analogous to both the free market and evolution. It allows for creative destruction, which is necessary to prevent stagnation.

The selection method is critical. It must be based on norms of rationality. If ideas are selected on some other basis, such as popularity, they will reflect that basis.

One of the problems with the academy is that it is largely self-governing, despite not being self-funded. Imagine if a company hired someone and said “Your job is whatever you define it to be, and we will pay you for whatever you do”. That is essentially the relationship between society and the academy.

An academic field should face oversight by people completely outside that field, who are not invested in the field’s theories. They would judge it only with respect to norms of rational thought and discourse. There could even be mechanical oversight by computer programs that are designed to identify fallacies in academic research.

However, there is always the problem of who watches the watchmen and who regulates the regulators. Any institution can become corrupt. It could be that we simply need to burn institutions to the ground periodically, and rebuild them.

What is the current purpose of the academy?

The social function of the academy is to generate, maintain and distribute general knowledge.

We can view the academy in other ways. We can view it as a self-perpetuating system, whose purpose is simply to perpetuate itself. Some properties of the academy have been selected to have that effect. Or we can view the academy as a pathology of abundance, analogous to obesity. Our civilization consumes a vast amount of energy from fossil fuels, and that energy must go somewhere.

We can also view the academy from an individual perspective. What does the individual get from a degree? The main benefit of a degree is status. To some extent, higher education is a status competition that does not benefit society as a whole. It is yet another tragedy of the commons.

You pointed out how assumptions like the ‘blank-slate’ view can influence research. What other unexamined assumptions do you think are currently undermining research?

The blank-slate assumption is a special case of the moralistic fallacy, which is a huge problem within the academy. The implications of the theory of evolution for human nature and society are widely ignored or denied, because they conflict with moral/religious assumptions.

There is a tacit assumption that rational inquiry cannot “rock the boat” of ordinary assumptions and intuitions. Most academics tacitly assume that their upper-middle-class values and left-wing political views are correct and rational. Thus, they assume that science and philosophy are compatible with those views. If there is a conflict, they ignore it or deny that it exists.

For example, there are academic philosophers who reject the objectivity of moral values and obligations, and yet continue to make moral claims and arguments. They take take their moral intuitions for granted as correct and rational, despite having exposed them as a delusion. They assume that there must be some other rational basis for moral claims and judgments.

Another example is Dawkins’ inability/unwillingness to understand/admit the implications of the theory of evolution for human beings. He does not commit the moralistic fallacy by assuming that our notions of moral goodness are manifest in nature. He accepts that life is competitive and selfish. But he does not take this train of thought to its logical conclusions: that our notions of moral goodness are fake, and that we cannot create a moral utopia. Instead, he simply gets off the train once it leaves the realm of biological science. He does not follow the track to its psychological, social and philosophical implications.

Map-territory conflation is another deep conceptual problem within the academy. In many fields, there is a tendency to slowly shift the focus of inquiry inward, so that the field becomes circular.

For example, in academic philosophy, the focus has shifted from philosophical problems to the works of past philosophers, such as Hegel, Kant, etc. The task of philosophy is to define and solve philosophical problems, not to study and critique philosophical texts. The latter could be relevant to the former, but it should not be the primary focus.

If universities often prioritize publications over real knowledge, what internal changes could they make to better evaluate the true intellectual contribution of research?

There is no perfect way to evaluate the intellectual contribution of research, and it certainly can’t be quantified. So, there is no point trying to do that. However, there are ways to assess the rationality of a field, research methods, and specific research. There are also ways to promote intellectual progress.

A single university can’t do much by itself, because it is part of a bigger social institution, the academy.

As a social institution, the academy should have a constitution: a formal statement of its social functions and its internal principles. That constitution should include an explicit commitment to rational inquiry and discourse. There should also be explicit standards of rational inquiry that every academic field is required to satisfy.

Academic fields should not be free to define their own methods and standards. There should be a separate agency within the academy that evaluates academic fields and research for rationality.

The academy should make sure that there is competition within fields, to prevent circular, self-justifying discourse games from emerging by tacit cooperation.

The academy should be a place for inquiry, not political advocacy. Pure advocacy fields, such as gender studies and ethnic studies, should be eliminated. Other fields, such as communications, sociology, political science, English, and the fine arts, should be significantly reformed.

There should be a division between research and teaching. Those functions should be done (mostly) by different people. There should be no requirement for teachers to publish research, or for researchers to teach.

Within scientific fields, there should be a distinction between core theoretical research and detailed research. Whenever possible, the dependence of detailed research on theories should be minimized. This modularity would help to prevent theoretical inertia, by making it easier to change core theories.

The academy should create and maintain an encyclopedia: a central repository of knowledge. (This is an example of how the internet has advanced beyond the academy.) This encyclopedia should allow for competing theories to be presented and compared. It should also allow for public comment and discussion, within limits.

As a public institution, the academy should be open and responsive to the public. Academic researchers should be required to engage in public discourse to some extent.

These reforms would make the academy a more functional institution.

If what we observe is already shaped by our theories, how can we ever truly test those theories with new observations without just confirming what we already believe?

If by “truly test” you mean “test in a purely objective way”, that is impossible. We can’t compare theories to reality. We can only compare predictions to observations, or assess how well a theory can explain observations.

Observations depend on both theories and human judgment. It does not follow that they are empirically vacuous. They also depend on reality.

Consider temperature, for example. The concept of temperature is both a theoretical construct and an ordinary, intuitive concept. We feel hot and cold. The two concepts are correlated, but not the same. At room temperature, a metal spoon feels colder than a wooden spoon. They have the same theoretical temperature, but different intuitive temperatures. The theory of heat, which includes the theoretical concept of temperature, can explain those experiences. (Metal is a better conductor of heat.)

A measurement of temperature with a thermometer is an observation. It depends on the theory of heat, because we use the theory to interpret what we see (millimeters of mercury in a glass tube). The measurement also depends on reality (the expansion or contraction of mercury). And it depends on human judgment and action (someone has to make the measurement, interpret it as a measurement of temperature, etc.).

If you observe your grandmother, that observation depends on your conceptual knowledge (your concept of your grandmother) and your brain. Does this dependence make the observation vacuous? No, of course not, because the observation also depends on reality. The same is true for scientific observations.

For more about knowledge, see Theories of Knowledge.

If academic research is just about solving “little fake problems,” how can we tell that apart from a legitimate, narrowly-focused inquiry that contributes incrementally to our understanding?

There is no magic oracle, but there are some heuristics.

First, don’t assume that every research problem is real. Start with with the default assumption that it isn’t, and then see if there is sufficient evidence to override that assumption.

Second, consider whether a field has any need for detailed research. Start with the default assumption that it doesn’t, and then see if there are good reasons to override that assumption.

For example, do we need thousands of mathematicians doing research in mathematics? Why? What is the point of it? There is no need to fill up journals with obscure mathematical results that no one will ever use. Even if someone needs a particular result, he will probably have to derive it himself, rather than wade through the literature.

Third, the further research is removed from core theory, the closer it should be to some practical application. If research is detailed but not practical, it is probably useless.

Again, some fields require the collection of many little observations. We should define the scope of such work before doing it. Do we want a complete inventory of all types of life on the planet? If so, then there is work to be done in that area. Do we want biologists to inventory the species in every ecosystem, their population sizes, their relationships, etc? Probably not. Instead of treating every question of knowledge as worth answering, we should prioritize research questions based on their potential for theoretical and practical impact.

Research problems are often justified by an appeal to the status quo. If academics currently do X, it is assumed that X has social value. But this is just a circle. Research problems should be justified with respect to theoretical and practical relevance.

If the “modern academy is not a reliable source of knowledge,” what are the long-term consequences for a society that continues to rely upon such an institution for education, innovation, and policy advice?

The society will be stupid.

As AI gets better, is it more likely to create more fake research or help us detecting them?

In the short term, I wouldn’t expect either effect.

An LLM is basically just Wikipedia that can talk to you. It only contains knowledge that humans have generated.

LLMs might create more cultural inertia, if people treat them as oracles. There is a danger that LLMs will be used to validate existing knowledge (fake or real), even though they were trained on that knowledge. This is another potential circularity to worry about.

We might be able to develop software that identifies errors of reasoning in academic research. Ideally, there would be a formal representation scheme for knowledge and justification. It would be something like formal logic, but more powerful, because it would cover the definitions of concepts, inductive generalization, empirical arguments involving data, etc. If we had such a representation scheme, researchers could encode their research in that scheme, or a machine could translate academic verbiage into that scheme. The output could then be evaluted mechanically. Machine review could eliminate some of the problems with peer review.

However, we don’t have that technology now. LLMs are not sufficient. We need “artificial rationality”. Maybe we will be able to automate even the creative aspects of knowledge production someday, but we are not there yet.

To what extent do you think the increasing emphasis on metrics, rankings, and quantifiable outputs in academic evaluations contributes to fake research?

I’m not aware of this change. I don’t see how academic output can be usefully quantified.

Goodhart’s law applies to any metric of academic achievement. For example, the number of papers published might once have been a good measure of the importance of an academic. But when academics know that they are being evaluated in that way, the metric becomes less meaningful.

When someone sees academic dishonesty, what obstacles stop them from telling research funders about it, and what kind of help from their institution would make it easier for them to speak up?

This question makes me wonder if Alex (or whoever made up these questions) actually read my essay.

The essay was not about academic fraud, such as plagiarism or making up data. That happens, but it is a small part of a much bigger problem. Most research is fake in other ways.

Here is an analogy. Suppose that I create a list of the cracks in the sidewalk on my street. Every day, I go out, find some cracks, measure them carefully, and record this information in a document. Am I producing real knowledge? It is accurate information about the world, so in that sense it is real knowledge. But it is completely irrelevant and uninteresting. In that sense, it is fake knowledge.

Most academic research is fake in that way. Sometimes, it contains logical errors or other mistakes. Sometimes, it is entirely made up. But even correct information can be fake knowledge.

Why do institutions fund fake research? That’s an interesting question.

Governments provide a lot of funding for academic research. This is yet another wealth redistribution scheme. It creates a class of people who depend on the government. It also creates a bureacracy to administer the funds. The government can claim that this wealth redistribution has social benefits, but that is just an assumption.

Can a funder measure the return on investment in research? In some limited cases, maybe. In most cases, no. What is the return on all the research into climate change? String theory? Cancer? There is no way to quantify it.

Can a funder select research (in advance) that is likely to generate a return? Again, in most cases, no. That’s the nature of research. It is search. You don’t know what you are going to find before you find it.

However, we can prune away potential research that is unlikely to generate a return. Most detailed research can be eliminated on the grounds that it is unlikely to have any theoretical or practical impact.

If a company has a specific industrial problem, it can invest in research to solve that specific problem. It can’t know in advance whether the research will pay off, but it can estimate the benefit of the research after the fact.

For general research, neither is possible. Again, research is search. It is speculative.

I’m not opposed to investing some money on general research. However, it is naive to assume that all the money spent on research is socially beneficial.

What can big research funding groups do to actually get researchers to be more honest, besides just cracking down on them when they mess up?

Again, this seems to completely miss the point. Alex must not understand my views on the academy at all. The problem is not individual researchers “messing up”. The problem is systemic.

The biggest research funder is the government. Government bureaucrats don’t care very much about the benefits of research. In most cases, the benefits can’t be measured, and the bureaucrats are not responsible for the results. So, they assign funds in a way that is conventional.

Even within companies that fund research, the incentives of the person assigning funds are not necessarily aligned with either company profits or the generation of knowledge.

There is no perfect solution to the problem of misaligned incentives. Institutions often become corrupt and inefficient. But there are partial solutions: explicit rules and accountability.

There are some other ways to improve the efficiency of the academy. Rather than funding research in advance, there should be more emphasis on rewarding people for results. It should be possible for people outside the formal institution to submit their original research (in a specified format) for consideration.

What are the biggest long-term problems for society if important government decisions are based on our current fake academic research?

Stagnation, decline, collapse.