Fabian Hundertmark, Bielefeld University
Jakob Roloff, Justus-Liebig-University Gießen
Francesca Bellazzi, University of Oslo, ERC Project Assembling Life (no.101089326)
1. Introduction
In the target post, Dong and Piccinini criticize SE and propose a new goal-contribution account of functions (GCA) (also in Maley, Piccinini 2017; Piccinini, 2020). In this second post, we argue that GCA suffers from several problems because it ignores the role of selection. Finally, we propose a compromise between SE and GCA while upholding that selection remains inescapable.
2. Past Selection is Inescapable
According to the traditional goal-contribution account by Boorse (1976), the function of a trait token is a statistically normal contribution of tokens of the same type to the goals of goal-directed systems; essentially, a common beneficial effect. GCA modifies this account in two ways. First, it substitutes the system’s goals with objective biological goals that populations must pursue to continue existing (e.g., survival, development, reproduction, and helping). Second, it specifies that functions must contribute to these goals reliably by being performed at appropriate rates under appropriate conditions.
According to GCA, the difference between a randomly beneficial effect and a function is that a function contributes to goals more regularly and reliably. Rita Marley’s dreadlocks may have had the beneficial effect of slowing down a bullet fired at her, saving her life. Nonetheless, her hair does not have the function of slowing down bullets because hair generally has this effect too infrequently and unreliably.
With this background, consider Dong and Piccinini’s claim that GCA can ascribe functions to “clearly functional” traits such as novel mutations and new artifacts. It is entirely unclear how GCA could ascribe functions to the first token of a type or at least the first performances until the performances reach the required regularity threshold. In the example above, would the first slower-beating heart have a GCA function? Although this claim is part of the alleged benefits of GCA, it lacks justification.
GCA faces even more problems. The first is the well-known problem of pandemic dysfunctions (Neander 1983, 83ff.; 1991, 182f.). Diseases, such as pneumonia, may prevent some trait tokens from performing their function reliably. In these cases, GCA generally and correctly attributes systematic malfunctions, or ‘dysfunctions’ as we call them. As a disease becomes more widespread, fewer traits can perform their function at the appropriate rate and circumstances. Until, at some point, the corresponding effect no longer occurs regularly enough to count as a function according to GCA. Consequently, the traits can also no longer be dysfunctional.
A related second worry is the problem of pandemic futile functioning. Most traits contribute to the attainment of biological goals only together with other traits or fitting environmental conditions. The eyes’ light conversion into electrical signals contributes to survival and other biological goals only if the rest of the organism is sufficiently healthy and the environment is sufficiently favorable. Consequently, pandemic diseases that affect other parts of the organism or far-reaching environmental changes can also cause traits to lose their function.
A third and more fundamental problem for GCA arises because the required frequency and reliability for functionality plausibly depend on the traits and goals in question. Let us contrast two cases. Sperm cells have the function of fertilizing mature eggs, although at most one sperm out of 300 million will fertilize an egg (example from Millikan 1984, 3), so the function is realized quite rarely. On the other hand, heart sounds are often used for diagnostic purposes and thus contribute to survival, but making sounds is not considered a function. In general, the question arises where to draw the line.
The reason for these three problems is that GCA requires functions to be performed and contribute to biological goals in the present. An obvious way to solve the first two problems is to allow activities to be functions because they have contributed to biological goals regularly and reliably in the past. The third problem can be solved by setting the successful selection of the trait as the ultimate biological goal. This is because the contribution of a trait to subordinate biological goals must have been regular and reliable enough to ensure past selectional success by being significant enough to outweigh a trait’s costs and adverse effects. Since, for example, reproduction is indispensable for (natural) selection, the relatively low reliability of sperm is still sufficient for them to be positively selected. Hair, on the other hand, was not selected for slowing down bullets, as there was no corresponding selection pressure.
To sum up, GCA’s problems can be solved by appealing to past selection processes. In other words, past selection is inescapable!
3. Conclusion: Rethinking functions
We have now shown that when rethinking functions, past selection processes should be considered (Part 2) and that doing so is not problematic (Part 1).
A final point is whether the pragmatic use of function ascription in science only refers to selected functions. There are good reasons to be skeptical (Ratti and Germain, 2022). Function ascriptions in the sciences may serve different explanatory purposes and occur in different practices. As philosophers, we should not be the “function police,” ensuring that all function attributions are compatible with the SE or any chosen theory of function. This would be impossible and probably not advisable either. We believe that the word “function” is often used as a synonym for “activity” or “effect” (Neander 2017, 52ff.), and accordingly, it can refer to a variety of different phenomena.
With this pluralism in mind, we want to end by proposing the blueprint account of functions that brings together the intuitions of proponents of the causal role, organizational, and goal-contribution accounts. The proposal starts by considering fully functional systems in their natural environment, with past selection processes determining whether a system is fully functional and whether an environment is natural. A blueprint function of a trait is a contribution that the trait would make to the capacities of the system if this system were fully functional and in its natural environment. Blueprint functions are not directly determined by past selection, since the parts need not have undergone their own selection processes.
Functions rethought in this way are consistent with large parts of scientific practice and many intuitions. This explains why scientists and even proponents of the SE theory rarely do evolutionary deep dives when attributing functions, as one of the commenters on Dong and Piccinini’s blog post rightly noted. At the same time, selection is fundamental in allowing for the identity, history, and context relativity of those bearing blueprint functions. So, do we have to re-think functions? Maybe, but however we think about them, it is clear once again that reference to past selection processes is inescapable.
Acknowledgments
The authors thank Peter Schulte, James Turner and Maximilian Lipski for their insightful comments on this piece. This article was written thanks to the collaborations promoted by the Functions in Philosophy Network. Get in touch with the authors if you would like to know more.
References
Maley, C. J., and Piccinini, G. (2017). A Unified Mechanistic Account of Teleological Functions for Psychology and Neuroscience. In D. M. Kaplan (Ed.), Explanation and Integration in Mind and Brain Science. Oxford University Press.
Neander, K. (1983). Abnormal Psychobiology. Ph.D. dissertation, La Trobe.
Neander, K. (2017). A Mark of the Mental: A Defence of Informational Teleosemantics. Cambridge, USA: MIT Press.
Piccinini, G. (2020). Neurocognitive mechanisms: Explaining biological cognition. Oxford University Press. https://doi.org/10.1093/oso/9780198866282.001.0001
Ratti, E. & Germain, P. (2022). A relic of design: against proper functions in biology. Biology and Philosophy 37 (4):1-28. https://doi.org/10.1007/s10539-022-09856-z