TO ERR IS (NOT ONLY) HUMAN

Concepts such as ‘purpose’ and ‘goal-directedness’ have an uneasy history in the philosophy of biology. They are teleological concepts, and teleology harks back to a pre-modern, quasi-Aristotelian understanding of biology. Biologists themselves use teleological language liberally, yet philosophers tend to be warier, perennially looking for ways to ‘naturalize’ teleology. And this means, roughly, explaining it without invoking the ‘spooky metaphysics’ of vitalism, or the ‘final causes’ that are supposed to exert their power from the future to the present, or the invisible hand of an intelligent mind directing organisms’ behaviour.

It is, however, possible to analyse the goal-directed, purposeful behaviour of living things in ways that are both amenable to empirical investigation and philosophically plausible without scaring the horses, as it were. Indeed, the global multi-disciplinary Templeton Foundation research programme Agency, Directionality, and Function, of which our team is a part, has this aim in mind. Our own line of research concerns the concept of ‘mistake-making’. We think this phenomenon is a universal feature of organisms, species, and even many parts and sub-systems of organisms. It is not only humans who make mistakes. The cat that jumps on a shrew thinking it’s a mouse and then spits it out, the domestic hen that tries to hatch a golf ball, the fish that takes the bait, the frog whose darting tongue misses the insect by millimetres: these are all examples of organisms getting things wrong—in other words, making a mistake. It is clear that what they were trying to do is not what they ended up doing, in a minimal sense of ‘trying’ that does not necessarily require effort or self-reflection. And what they ended up doing is not the sort of thing they are supposed to be doing as far as their own intrinsic natures are concerned. That, broadly, is what we mean by making a mistake.

Think of the liberal use of mistake-related terminology even at the micro-scale and at the sub-systemic level. Bacteria can be confused by plants mimicking signals used by the bacteria for infection of the plant (Bauer and Mathesius [2004]). Antibodies can be fooled by pathogens. Competitive inhibitors can trick an enzyme into binding to the incorrect substrate. Just how close to universal mistake-making is in biology is an exciting project for biologists to consider. We hypothesize it might even be found in blood platelets if they can be fooled by collagen-like proteins (instead of collagen protein itself) into beginning a blood clotting process that is not supposed to happen (and could even be life-threatening).

Philosophically, it is vital to distinguish mistakes in general from the mistakes typically made by humans. Such mistakes are often conscious or can be brought to consciousness through reflection and involve moral concepts such as carelessness and recklessness, as well as issues to do with free will, responsibility, accountability, and so on. We claim it is possible to work our way outwards from what is typically human to an extended concept of biological mistakes in general, of which human mistake-making is a distinctive instance.

Moreover, what we find of particular importance in the concept of biological mistakes is a way of operationalizing teleology, namely, making it the subject of empirical investigation, turning it into material for the generation of novel, testable hypotheses of interest to the working biologist. We set out some of these ideas in a recent paper (Hill et al. [2022]) and develop them in our current BJPS article, providing several examples. For instance, researchers have described working on cell ‘decision-making errors’ resulting from noise and signalling failures (Habibi et al. [2017]). They found that a cell can respond differently to the same input, leading to incorrect cell decisions and responses. One kind of mistake is a ‘false alarm event’, where there is no object (such as a particular protein) but noise leads the system erroneously to declare the presence of that object; another is a ‘miss event’, where the system misses the presence of the object. The likelihood of a false alarm or miss event can be computed using probability distributions, but our mistakes framework leads us to ask the following questions: First, is there a cellular mechanism for distinguishing signal and noise, thus enabling direct investigation rather than indirect via probability distributions? And, second, can cells mismeasure the amount of a given protein in this kind of decision-making situation and, if so, what is the underlying mechanism?

We do not suppose that, for a given hypothesis involving mistake-making, a biologist could or would not have come up with it without appealing explicitly to the theory of biological mistakes. Rather, our theory is a way to organize many of the thoughts and concepts experimentalists naturally work with when they generate hypotheses for testing. But also, we hope, the theory of biological mistakes will encourage biologists to consider phenomena that may not have previously occurred to them. What mistakes does their target organism typically make? What potential might it have to make other kinds of mistake? What underlying mechanism might there be by which a certain kind of mistake is made, and how common is it (if at all) across a variety of organisms of widely different kinds and at different scales? How does their organism avoid mistakes or correct them? And if it cannot, why not? How does mistake-making vary with the kinds of trade-off an organism may make between, say, timely action in its environment and sufficient or effective action (say, to avoid a predator, or find a mate, or repair damage)?

It is exciting that the mistakes framework is capable of generating many ways—including, we believe, novel ways—of understanding and testing a system. That said, difficult conceptual questions need to be resolved in order to firm up our framework. For example, there is a broad sense of ‘mistake’ that encompasses what we investigate in our present BJPS article. Included in this would be all states, conditions, structures, processes, and so on that deviate from what we call a ‘standard of correctness’ for a system—say, by undermining an organism’s health, or bodily integrity, or survival. However, we focus on the making of mistakes. Strictly, every mistake is made, and this requires a concept of agency by which the behaviour of a system (organism, or part, or cell, or species, or group) can be said to amount to the making of a mistake. There is, we claim, a metaphysical difference between mistake and mere failure. It is a mere failure to be hit on the head by an unexpectedly thrown rock. It is a stupid mistake to try to eat a rock! Mistakes are things we do; failures are what simply happen to us. We develop a rigorous concept of what we call the ‘minimal biological agent’ to accommodate this need—a concept that applies as much to the actions of enzymes as to those of insects or birds.

We also need, of course, a rigorous definition of biological mistake, and this we propose as well. Our basic idea here is that a mistake is something done by an organism (or similar) that threatens its effective action in its environment. We adopt a liberal concept of threat, according to which an act is a threat if, when not mitigated, it undermines further effective action (or ‘function’ in a broad sense). A deer that walks in front of an oncoming car is subject to a threat: if the car doesn’t stop or the deer doesn’t move fast enough, it will be injured or killed. A person who ingests ethylene glycol has made a potentially fatal mistake, but so has the alcohol dehydrogenase enzyme that binds to it, since the enzyme’s binding action is a threat to the organism to which it belongs. This threat can be mitigated by introducing ethanol into the system, which is a competitive inhibitor to ethylene glycol; it reduces or removes the toxicity. But when the enzyme binds to the ethanol in this situation, there is no threat to the system—presumably the medical worker mitigates it by controlling how much ethanol itself is introduced. It would, however, be a mistake for the enzyme to bind ad libitum to ethanol, as the effect would also be toxic.

Yet an enzyme has no awareness of what it does; it just acts, and the relevant standard of correctness is that enzymes act for the benefit of the organism to which they belong, according to that organism’s specific needs. When they fail to do so, they err—yet without necessarily malfunctioning. Can enzymes be trained or conditioned to bind only to certain substrates (for the system’s benefit) and not others (that will lead to harm), even if the geometry is perfect for both? If binding to the wrong substrate is a case of being fooled, how do enzymes protect themselves from binding to incorrect substrates willy nilly, given that exposure to them is likely to be common in many typical situations? We think mistake theory can provide a structure within which these sorts of question can be systemically investigated.