The mismeasure of machine: why machine metaphors in biology are misleading

Time to indulge in the occasional revisiting of one of my technical papers, in the hope that they may be of more general interest then the original audience they were written for. This time I’m going to focus on one that I co-wrote with my long-time collaborator, Maarten Boudry, and published in 2013 in the journal Studies in History and Philosophy of Biological and Biomedical Sciences. The title of the paper is: “The mismeasure of machine: synthetic biology and the trouble with engineering metaphors.”

We began by noting that the scientific study of living organisms is permeated by machine and design metaphors. Genes are often characterized as providing the “blueprint” for an organism, organisms in turn are “reverse engineered” to discover their functionality, and living cells are compared to biochemical factories, complete with assembly lines, transport systems, messenger circuits, etc.

Although the notion of design is indispensable to think about adaptations, and engineering analogies have considerable heuristic value (e.g., when it comes to deploying optimality assumptions), Maarten and I argue in the paper that they are limited in several important respects. In particular, the analogy with human-made machines falters when we move down to the level of molecular biology and genetics.

Living organisms are far more messy and less transparent than human-made machines, as David Hume famously put it when addressing the classical argument from design. In part II of Dialogues Concerning Natural Religion, he writes:

“If we see a house … we conclude, with the greatest certainty, that it had an architect or builder because this is precisely that species of effect which we have experienced to proceed from that species of cause. But surely you will not affirm that the universe bears such a resemblance to a house that we can with the same certainty infer a similar cause, or that the analogy is here entire and perfect.”

Indeed, Hume continued:

“A continual circulation of matter in [the universe] produces no disorder; a continual waste in every part is incessantly repaired: The closest sympathy is perceived throughout the entire system: And each part or member, in performing its proper offices, operates both to its own preservation and to that of the whole. The world, therefore, I infer, is an animal.”

So, if anything, the universe resembles the messiness and organic development of living beings, not the precision and exact functionality of machines. The analogy with man-made artifacts, he concludes, is flawed and should be rejected. If that’s true at the level of the cosmos, we think it is also true — and for similar reasons — at the level of biological organisms.

A better way to think of evolution is as an opportunistic tinkerer, blindly stumbling on “designs” that no sensible engineer would come up with. This was pointed out by Francois Jacob back in his classical 1977 paper, “Evolution and Tinkering,” where he says:

“The action of natural selection has often been compared to that of an engineer. This, however, does not seem to be a suitable comparison. First, because in contrast to what occurs in evolution, the engineer works according to a preconceived plan … Second, because of the way the engineer works: to make a new product, he has at his disposal both materials specially prepared to that end and machines designed solely for that task. Finally, because the objects produced by the engineer … approach the level of perfection made possible by the technology of the time. In contrast, evolution is far from perfection. … Natural selection has no analogy with any aspect of human behavior. … It works like a tinkerer — a tinkerer who does not know exactly what he is going to produce but uses whatever he finds around him whether it be pieces of string, fragments of wood, or old cardboards; in short it works like a tinkerer who uses everything at his disposal to produce some kind of workable object.”

That is, natural selection is a satisficying — not optimizing — process, as well as the ultimate recycler!

If you are thinking that perhaps we are building a strawman, that nobody really thinks of organisms as engineered, here is George Williams, one of the foremost evolutionists of the second part of 20th century:

“Whenever I believe that an effect is produced as the function of an adaptation perfected by natural selection to serve that function, I will use terms appropriate to human artifice and conscious design. The designation of something as the means or mechanism for a certain goal or function or purpose will imply that the machinery involved was fashioned by natural selection for the goal attributed to it.”

And yet, even arch-adaptationist Richard Dawkins — whose popular work was derived in part from William’s — had this to say while watching the dissection of a giraffe’s neck:

“Not only would a designer never have made a mistake like that nervous detour; a decent designer would never have perpetrated anything of the shambles that is the criss-crossing maze of arteries, veins, nerves, intestines, wads of fat and muscle, mesenteries and more.”

Maarten and I also point out that the engineering-inspired idea that natural selection is capable of “solving a near intractable physics-problem,” as Steven Pinker said with regard to the smooth movement of your limbs, though having a kernel of truth, is profoundly misleading. Animals don’t use algebraic fractions to calculate the level of altruism they should extend to their kin (not even unconsciously), any more than birds use latitude and trigonometry to navigate to their brooding places, or dogs compute parabolic trajectories when they’re catching a ball in flight. All these animals use surprisingly simple rules of thumb which, in their specific ecological environments, produce behaviors that more or less track engineering solutions.

Yet in popular science books the language may be ambiguous. Dawkins, for instance, writes in The Selfish Gene:

“When a man throws a ball high in the air and catches it again, he behaves as if he had solved a set of differential equations in predicting the trajectory of the ball. He may neither know nor care what a differential equation is, but this does not affect his skill with the ball. At some subconscious level, something functionally equivalent to the mathematical calculations is going on.”

Well, much hinges on what one means by “functionally equivalent.” Experiments show that humans (and dogs) use a deceptively simple heuristic to catch a ball: keep your gaze fixed at the ball, and adjust your running speed such that the angle of the ball remains constant (for references to this and other claims in this post, see the original paper). When you follow this heuristic, you will be there when the ball hits the ground. As it happens, baseball players are very poor at predicting where a ball is going to hit the ground when they are asked not to run towards it. They just manage to get there when the ball does. This is a little surprising since computing the trajectory of a ball is a very complicated physical problem: one has to take into account initial velocity, angle, direction, spin, as well as the air current and the distance from the player.

We then move on from a preliminary discussion of the use of engineering metaphors in biology to consider more directly the field of synthetic biology. To begin with, there is no such thing as a single research program in this emerging area. The literature distinguishes at least five conceptually distinct, if somewhat overlapping programs associated with synthetic biology:

(1) Bioengineering. Uses standard biotechnology tools to build novel biochemical pathways in host organisms.

(2) In silico synthetic biology. Similar to bioengineering, but carried out using computer simulations of novel metabolic pathways, rather than by experimentation with living organisms.

(3) Synthetic genomics. As the name plainly implies, this is a much broader scale of bioengineering intervention, at the level of whole genomes — rather than individual pathways — being slated into a (de-genomicized) host cell.

(4) Protocell synthetic biology. Here the aim is somewhat complementary to that of synthetic genomics: to bioengineer “living” cells that could then be used as entirely artificial hosts for other bioengineering projects.

(5) Unnatural molecular biology. This approach is arguably the most ambitious, as researchers in this area pursue the goal of producing entirely new molecular biologies, for instance using expanded genetic codes, capable of incorporating more and different amino acids from those used by the natural code.

Despite impressive technological innovation, the prospect of artificially designing new life forms from scratch has proven more difficult than the superficial analogy with “programming” the right “software” might have initially suggested. The idea of applying straightforward engineering approaches to living systems and their genomes — isolating functional components, designing new parts from scratch, recombining and assembling them into novel life forms — pushes the analogy with human artifacts beyond its limits and onto the breaking point. In the absence of a one-to-one correspondence between genotype and phenotype (which does hold, instead, in the case of blueprints and actual engineering projects), there is no straightforward way to implement novel biological functions and design new life forms.

Both the developmental complexity of gene expression and the multifarious interactions of genes and environments are serious obstacles for “engineering” a particular phenotype. The problem of reverse-engineering a desired phenotype to its genetic “instructions,” we suggest, is probably intractable for any but the most simple phenotypes, and recent developments in the field of bio-engineering and synthetic biology reflect these limitations.

Instead of genetically engineering a desired trait from scratch, as the machine/engineering metaphor promises, we suggest that researchers are more likely to make progress by co-opting natural selection itself to “search” for a suitable genotype, or by borrowing and recombining genetic material from extant life forms.

Maarten and I conclude the paper by suggesting that perhaps we should be looking for new metaphors, or even shy away from metaphorical language whenever possible.  One alternative metaphor for thinking about the relationship between genomes and phenomes is the idea of a recipe, where DNA contributes the equivalent of the instructions for cooking, but does not specify all of the details of the process, which are left to a continuous interaction between the recipe itself and the environment and ingredients that are being used.

Although the recipe metaphor does get us away from a straightforward talk of “blueprints,” and particularly from a simplistic, near one-to-one Genotype => Phenotype mapping function, it is of mostly educational use and is unlikely to generate novel insights to guide professional researchers.

The same holds another common metaphor, that of an origami, proposed by Lewis Wolpert. It captures some important elements of embryological development (like the circuitous step-by-step folding), but it obviously will not work as a new master metaphor for thinking about living organisms (nor was it intended as such).

While we acknowledge that metaphorical and analogical thinking are part and parcel of the way human beings make sense of the world, in some highly specialized areas of human endeavor it may simply be the case that the object of study becomes so remote from everyday experience that analogies begin to do more harm than good (Hume docet). In particular, the systematic application of engineering metaphors to a domain that is fundamentally different from the world of human artifacts may send scientists on a wild goose chase. Wittgenstein famously said that “Philosophy is a battle against the bewitchment of our intelligence by means of our language.” Perhaps a contribution of philosophy of biology to the field of synthetic biology is to help free the scientists from the bewitching effects of misleading metaphors, so that they can simply get on with the difficult and unpredictably creative work lying ahead.


P.S.: a few sentences in this essay have been edited to reflect critical points raised by some readers, see below.


249 thoughts on “The mismeasure of machine: why machine metaphors in biology are misleading

  1. Robin, I out-thunk you. The bicycles are at my house. You’ve been Bright-sided. 😉 … My mind is also beaming Nigel Farange over to Coel’s to read out loud to him every comment we’ve made here. “We hafffff wayyys of making you be silent, Mr. Maxwell Schmahhht!” (Apologies to those who have not seen 1960s American TV.)


  2. From the standpoint of a consumer of science writing I think the fewer metaphors and analogies the better, although that is just a personal preference. I just like a plain, clear description of what is going on. I have read a few good books in this respect – usually they are much easier to understand and cover more ground than the ones that try to be fancy.


  3. Hi marc lavesque,

    I agree with most of your comment, and the different functionally equivalent programs is a good comparison. But on this bit:

    though the output ‘the ball is caught’ is the same in each case the input isn’t, it’s ongoing: “keep your gaze fixed at the ball, and adjust your running speed such that the angle of the ball remains constant”.

    The input is surely: “I see a ball, I want to catch it”.

    The heuristic (“keep your gaze fixed at the ball, and adjust your running speed such that the angle of the ball remains constant”) is not the “input”, it is instead the program that the student writes to get to the output.

    The output, I agree, is then “the ball is caught”.

    Another student in the class might have programmed: “deduce the trajectory of the ball from visual information, compute landing place and time of ball, go there, clasp hand at that time”.

    If the two programs both arrive at the same “the ball is caught” output then — as you say — they are functionally equivalent.

    Liked by 1 person

  4. Morning Socratic,

    No, Coel, the fallacy is thinking that you would ever admit you’re wrong.

    I would be greatly aided in that matter if you would try making half-way sensible criticisms.


  5. Hi Robin,

    The one which “behaves as if” it were solving DEs will be that one that is, in fact solving DEs …

    No, you’re still wildly over-interpreting this whole thing. The “as if” language implies that it is not actually doing that thing, but something else that ends with the same output. Let’s continue from marc’s comment:

    Input: “I see a ball, I want to catch it”.

    Output: “the ball is caught”.

    Then A, B, C, D, E … are programs that the class write, some involving DEs and some heuristics, all of them arriving at the output “the ball is caught”.

    “As if” is short for “as would be the case if”. If the system uses A, then it arrives at the output “the ball is caught” as would be the case if it had used B, C, D or E.

    Similarly, if it were using B, then it arrives as the output as would be the case if it had used A, C, D etc.

    All of A, B, C, D, E are “functionally equivalent” in the sense of giving the output given the input.

    All of this is close to tautological as soon as one defines the intended “function” as catching the ball. And, as the writer, Dawkins was entitled to pick the “function” he was talking about.


  6. Coel,

    “The input is surely: “I see a ball, I want to catch it”.”

    I don’t think so. I think that comes before the input. Like the programming example, I don’t think the input is “I want to solve the teacher’s assignment”.


Comments are closed.