In a previous post I explored the decline of science as related to the decline capitalism. A large aspect of this decline is how the increase of informational complexity leads to marginal returns in knowledge. For example, the last revolution in physics appeared roughly one hundred years ago, with the advent of quantum mechanics and relativity. Since then, the number of scientists and fields have exponentially increased, and the division of labor has become increasingly more complex and specialized. Yet, that billion dollar per year experiment, the Large Hadron Collider, that was created to probe the most fundamental aspects of theoretical physics, has failed to confirm any of the new theories in particle physics. The decline of science is coupled to the decline of capitalism in general, as specialist and institutional overhead is increasing exponentially across industries, but GDP growth has been sluggish since the 1970s.
Right now across scientific fields there is an increasing concern for the overproduction of “bad science”. Recently, the medical and psychological sciences have been making headlines, because of the high rates of irreproducible papers. In even the more exact sciences, there is a stagnant informational bloat, with a flurry of math bubbles, theoretical particles, and cosmological models inundating the peer-review process, in spite of billion dollar experiments like the Large Hadron Collider not confirming any of them, with no scientific revolution (last one was 100 years ago) in the horizon.
There is no shortage of solutions being postulated to solve the perceived problem. Most of them are simply suggestions of making the peer review process more rigorous, and refining the statistical techniques used for analyzing data. For example using bayesian statistics instead of frequentism, encouraging the reproducibility of results, and finding ways to constraint the “p-value” hacking. Sometimes some writers that are a little bolder would argue that there should be “interdisciplinarity”, or that scientists should talk more to philosophers, but usually these calls for “thinking outside the box” are very vague and broad.
However, most of these suggestions would simply exacerbate the problem. I would argue that the bloat of degenerative informational complexity is not due to lax standards. To give an example, let’s analyze the concept of p-value hacking. A common heuristic in the social sciences is that for a result to be significant, it should have a p-value of less than 0.05. In layman parlance, this implies that your result has only 5 percent of probability of being due to chance (not exact definition but suffices for this example). So now you established a “standard” that can be gamed in the same way lawyers can game the law. This creates a perverse incentive to game this rule, by researchers finding all sorts of clever ways of “p-hacking” their data so that it passes that standard. So in the case of p-value hacking, one can make conscious fraud by not including the data that raises the p-value (high p-values mean your results are due to chance), to unconscious biases like ignoring certain data points because you convince yourself they are a measurement error, in order to protect your low and precious p-value.
The more rigid rules a system has, the more is invested in “overhead” to regulate those rules and game them. This is intuitively grasped almost by everyone, and hence the standard resentment against bureaucrats that take the roundabout and sluggish way to accomplish something. In the sciences, once a an important study/experiment/theorem generates a new rule, or “methodology”, this creates perverse incentive loops where scientists and researchers use this “rule” to create paper mills, that will in turn be used to game citation counts . Instead of earnest research, you have an overproduction of “bad science” that amounts the gaming of certain methodologies. String theory, which can be defined as a methodology, was established as the only game in town a couple of decades ago, which in turn constrained young theoretical physicists in investing their time and money in gaming that informational complexity, generating even more complexity. Something similar happens in the humanities, where a famous (usually french) guy establish a methodology or rule, and the anglo counter-parts game the rule to produce concatenations of polysyllabic words. Furthermore this fetish of informational complexity in the form of method and rules, creates a caste of “guild keepers” that are learned in these rules and accrue resources and money without allowing anybody who isn’t learned in these methodologies.
This article serves as a “microphysical” account of what leads to the degenerative informational complexity and diminishing returns I associated with modern science in my previous post. However what would be the solution to such a problem? The answer is in one word: ergodicity.
As said before, science has become more specialized, complex, and bloated that ever before. However, just because science has grown exponentially, it doesn’t mean it has become more ergodic. By ergodic I specifically mean that all possible states are explored by a system. For example a dice that is thrown a large amount of times would be ergodic, given that the system would access every possible side of the dice. Ergodicity has a long history in thermodynamics and statistical mechanics, where physicists often have to assume that a system has accessed all its possible states. This hypothesis allows physicists to calculate quantities like pressure or temperature by making some theoretical approximations of the number of states a system (e.g. a gas ) has. However we can use the concept of ergodicity to analyze social systems like “science” too.
If science were ergodic, it would explore all possible avenues of research, and individual scientists would switch of research programs frequently. Now, social systems cannot be perfectly ergodic, as social systems are dynamic and therefore the “number” of states grow (e.g. the number of scientists grow). But we can treat ergodicity as an idealized heuristic.
The modern world sells us ergodicity as a good thing. Often, systems describes themselves as ergodic as a defence from detractors. For example, when politicians and economists claim that capitalism is innovative, and that it allows all workers to have a chance at becoming rich (or a chance for rich people to become poor), they are implicitly describing an ergodic system. Innovation implies that entrepreneurs experiment and explore all possible market ideas so that they can discover the best ones. Similarly, social mobility implies that a person has a shot at becoming rich (or if already rich, becoming poor) if that person lives long enough. In real life, we know that the ergodic approximation is really poor for capitalism, as the rich do often stay rich, and the poor will stay poor. We also know that important technological innovation is often carried out by public institutions such as the american military, not the private sector. Still, the reason why ergodicity is invoked is that it is viscerally appealing. We often want “new blood” into fields and niches, and we resent bureaucrats and capitalists insulated from the chaos of the market for not giving other deserving people a chance.
One of the reasons that ergodicity is appealing is that there is really no recipe for innovation except experimentation and exploring many possible scenarios. That’s why often universities have unwritten rules of not hiring their own graduate students into faculty positions – they want “new blood” from other institutions. A common (although incorrect, as described above) argument against public institutions is that they are construed as often dull and stagnant in generating new products or technologies compared to the more “grassroots” and “ergodic” market. So I think there is a common intuition amongst both laymen and many professionals that the only sure way of finding if something “works” or not is trying different experimental scenarios.
Now let’s return to science. The benefit of ergodicity in science was indirectly supported the infamous philosopher Feyerabend. Before him, philosophers of science tried to come up with recipes of what works in science or not. An example is Popper, who argued that science must be falsifiable. Another example is Lakatos, who came up with heuristics of what causes research programs to degenerate. Yet, Feyerabend argued that the only real scientific method is that “anything goes” – he termed this attitude as epistemological anarchism. He argued that scientific breakthroughs don’t follow usually any hard and fast rules, and that scientists first and foremost are opportunists.
Feyerabend got a lot of flack for these statements – his detractors accusing him of relativism and anti-scientific attitudes. Feyerabend didn’t help himself because he often was inflammatory in purpose and seeking to cause a reaction (for example putting astrology and science on the same epistemic level). However I would say that in some sense he was protecting science from dogmatic scientists. To use the terminology sketched in the previous paragraphs: he ultimately was arguing for a more ergodic approach to science so that it doesn’t fall under this dogmatic trap.
This dogmatic trap was already explained in previous paragraphs: the idea that more methods, rules, divisions, thought policing, and rigour, would always lead to good science. Instead it leads to a growth of degenerative research that amounts to gaming certain rules. This in turn leads to the growth of degenerative specialists that are only experts in degenerative methods. Meanwhile, all this growth is non-ergodic, because it’s based around respecting certain rules and regulations, which constrains the exploration of all possible scenarios and states. It’s like loading a dice so that always the six dots face up, in contrast to allowing the dice to land in all possible states.
How can we translate these abstract heuristics of ergodicity into real scientific practice? The problem with much of philosophy of science, both made by professional philosophers, or professional scientists unconsciously doing philosophy, is that it looks at individual practice. It comes up with a laundry list of specific rules of thumb that an individual scientist most follow to make their work scientific, including certain statistical tests and reproducibility. However the problems are social and institutional, not individual.
What is the social and institutional solution? Proposing solutions is harder than describing the problem. However I always try to sketch a solution because I think criticism without proposing something is somewhat cowardly – you avoid opening yourself up to criticisms from readers.
The main heuristic for solving these problems should be on collapsing the informational complexity in a planned, transparent, and accountable way. As mentioned before, this informational complexity is like a cancer that increasingly grows, and its source is probably methodological dogmatism, where complex overhead becomes bloated as researchers find increasingly more convoluted way of “gaming” these rules. Here are some suggestions for collapsing complexity:
- Cutting administrative bloat and instead have rotating academics in the essential administrative postings.
- Get rid of the peer-review system, and instead use an open system, similar to Arxiv.
- Collapsing some of the academic departments into bigger ones. For example, there is more in common with much of theoretical physics, mathematics and philosophy than between theoretical physics and some of the more experimental aspects of physics. So the departments should be reorganized so that people with more similarities interact with each other.
- Create an egalitarian funding scheme, based more on divisions between theory and experiment than between departments. Everyone involved in the same category should receive the same, minimum amount of funding, where funding quantities are based on how much resources a specific type of work would realistically require. For example, a theoretical physicist that uses only pencil, paper, and their personal computer, has financially a lot in common with a sociologist that does the same.
- Beyond the minimum funding outlined above, excess funding should be decided democratically, with input outside of professionals.
- Abolish the distinction between tenured professor and adjunct. Instead everyone should teach.
Hopefully the destruction of admin bloat and adjunct/tenure distinction would release resources that can be spent on hiring researchers, instead of coming up with bad heuristics such as publication and citation numbers as filters for new hires.
Many of these recommendations cannot be seen in the abstract, since the University is intimately coupled to the society and the economy as a whole. For example, part of the admin bloat comes from legal liabilities and the state offshoring some of their responsibilities to universities. Number 6 would require a radical reconfiguration of society in general. Number 6 wouldn’t be able to be enacted today, since “democratic” institutions are composed of non-ergodic, technocratic lifers.
This takes me to the political conclusion that the problems of science should be seen as the problems of society as a whole. The only sure way to find solutions for problems is an ergodic approach. Right now, the state is non-ergodic, that is, its occupied and controlled by political and bureaucratic lifers. These non-ergodic bureaucracies in turn generate informational complexity, as new regulations and “rules” are imposed by the same caste of degenerative professionals, which in turn requires even more complex overhead. Instead, the State, (and in a socialist society, the means of production) should have a combination of democratic and sortition mechanisms that makes it impossible for individuals to stay too long in power. This democratic vision should be supported by broad and free education programs that train individuals with the sufficient knowledge required to rule themselves in a republican way. Not only is this method guarantees more equality, but it also turns society into this great parallelized computer that solves problems by ergodic trial and error, through the introduction of new blood, sortition and democratic accountability.
If you liked this post so much that you want to buy me a drink, you can pitch in some bucks to my Patreon.