Review of Paul Humphreys’s “Emergence: A Philosophical Account”
I am a cognitive scientist, and two of the most heated debates in cognitive science concern the nature of mind and the nature of consciousness. In recent years, almost identical questions have resurfaced around artificial intelligence. Does AI genuinely understand anything? Does it have a mind? Can it feel? Can it undergo experience? Could it be conscious? What kinds of cognitive capacities can we legitimately attribute to it?
In almost all of these discussions, one word appears with remarkable frequency: emergence. It is invoked so often in response to difficult questions that, at times, it seems less like an explanation than a way of declaring the problem solved. What is the mind? The mind is emergent. What is consciousness? Consciousness is emergent. Can phenomenal experience be explained? Of course, it is emergent. Could AI be conscious? Perhaps consciousness emerges in sufficiently complex systems. Could AI have emotions? Emotions are emergent, so why not? Is there genuine meaning in an artificial system? Meaning, too, is emergent.
And so it goes... Apparently I am not the only person bothered by this inflationary use of the term. The philosopher John Heil has also recently complained about the way emergent has become overused to the point of trivialization.
This was roughly the agenda I had in mind when I picked up Paul Humphreys’s Emergence: A Philosophical Account. I enjoyed the book immensely. I was impressed by Humphreys’s philosophical and argumentative precision, and I learned a great deal about the emergence literature along the way.
The book consists of seven chapters. In the first, Humphreys right away accepts that there may be no simple, a priori set of necessary and sufficient conditions that captures every legitimate case of emergence. Instead, he proposes a more indirect route. Emergence, he argues, often becomes an issue when a certain kind of reductive strategy, what he calls generative atomism, fails. If so, one promising way to understand emergence is to examine exactly where and how generative atomism fails. Humphreys therefore lays out the conditions under which generative atomism would succeed and then considers cases in which those conditions are not met.
He also distinguishes between ontological, inferential, and conceptual emergence. One of the points I found especially important here is something that is often overlooked in discussions of emergence, that is, the fact that a model of a phenomenon displays emergent behavior does not by itself warrant the conclusion that the phenomenon itself is emergent in an ontological sense.
With that distinction in place, the second chapter turns to ontological emergence. Humphreys departs from much of the literature by giving priority not to synchronic but to diachronic emergence. Temporality matters. Emergence is not simply a relation between simultaneously existing “levels,” but may involve genuine transformations unfolding over time. He develops this idea through two categories, transformational emergence and fusion emergence.
The third chapter turns to the historical background of emergence. Humphreys discusses the rise and decline of British Emergentism, as well as several concepts that have accompanied emergentist thinking: levels, downward causation, relationalism, holism, contextualism, and others. One difficulty quickly becomes apparent. Although these terms recur throughout the literature, different philosophers often mean substantially different things by them.
The fourth chapter is devoted to inferential emergence. Here, emergence concerns cases in which macro-level phenomena do not naturally follow from micro-level explanations. In some systems, for instance, the only way to determine what will happen is effectively to run or simulate the system step by step. In such cases, the emergence problem concerns the mathematical, computational, and epistemic resources available to us rather than necessarily reflecting a novel ontology. Humphreys makes an important distinction here. Derivability, predictability, and explainability are not the same thing. This matters especially when thinking about the relation between micro- and macro-level phenomena. A phenomenon may be difficult to derive without being inexplicable, it may be predictable without being reducible in a strong sense, or it may be retrospectively explainable even if it was not prospectively predictable. Another point that caught my attention in this chapter was his discussion, drawing on Mark Bedau, of weak and strong emergence. I had usually encountered the distinction in a fairly familiar form in that strong emergence refers to “spooky” cases such as consciousness, mind, or life, and weak emergence used for epistemological cases. Humphreys offers a more nuanced formulation, and the distinction becomes considerably less crude than it often appears in popular discussions.
The fifth chapter concerns conceptual emergence, and it was one of my favorite parts of the book. Sometimes, Humphreys suggests, the conceptual resources available within a given domain are insufficient for understanding a phenomenon, and we need to introduce a new conceptual vocabulary. In such cases, the relevant concepts can themselves be understood as emergent. Importantly, conceptual novelty is not presented as a separate and mysterious metaphysical category. More often, it accompanies ontological or inferential emergence as a need for redescription. Through discussions of reduction and construction, the Carnapian tradition, and functional reduction, Humphreys grants that many higher-level properties can be functionally characterized and connected to more basic realizers. But whether such functionalization succeeds is contingent. There may be no suitable lower-level realizer for a given property, or the conceptual transformation required to connect the two vocabularies may itself carry substantial explanatory information. For that reason, one cannot move directly from “we can describe this in lower-level terms” to “the higher-level concept is dispensable.”
The sixth chapter broadens the discussion to several philosophical issues closely associated with emergence. Supervenience, physicalism, fundamentality, multiple realizability, and related topics... Humphreys treats these debates with considerable care and resists the idea that emergence is automatically incompatible with physicalism. To say that something is emergent is not to say that it is ghostly, spooky, or metaphysically second-rate.
The seventh and final chapter examines scientific cases associated with emergence and asks how well the philosophical framework applies to them.
There is much more in the book than I can discuss here. What I appreciated most was the attempt to construct a genuinely careful and discriminating framework for emergence. Humphreys is perfectly willing to accept that emergence may be commonplace, that it can occur across many domains, without thereby making the concept trivial. Whether something is emergent remains an empirical and contingent question. We have to investigate the system in question. Thus there is still room for novelty and surprise, even though Humphreys rightly rejects defining emergence merely in terms of whether an observer happens to find something surprising.
I also agree with his suggestion that several puzzles surrounding emergence become easier to understand once we take diachronic emergence seriously. If anything, I think this aspect deserves even more emphasis. I would be inclined to go further and argue that there may be no metaphysically deep category of synchronic emergence at all. What we call synchronic emergence may often be a pragmatic consequence of the temporal or spatial resolution at which we choose to describe a phenomenon, a useful representational device rather than a fundamental metaphysical relation.
I was similarly struck by Humphreys’s skepticism toward the familiar language of levels. Levels-talk is often treated as if it were self-explanatory, yet it is surprisingly difficult to pin down. Humphreys suggests that talk of domains may often be preferable. I am not sure that I fully agree, but the proposal stayed with me.
At the same time, Humphreys explicitly avoids the more “sensational” cases that dominate contemporary discussions of emergence, namely, mind, consciousness, and life. He does not think these should be treated as the defining examples of emergence. His framework is therefore not presented as an explanation of consciousness, mentality, or life, even though he occasionally notes that there is no categorical reason why it could not eventually be applied to such cases.
This was, I have to admit, where I felt some disappointment. I find Humphreys’s account plausible. But many of the cases his framework handles are cases about which I am perfectly happy to say, “Fine, let us call that emergence. Or, for that matter, let us call it something else.” We could perhaps characterize many of these relations using a different vocabulary without losing much. On the other hand, what makes emergence philosophically urgent today, in my view, is not primarily the modest claim that complex systems sometimes exhibit novel organization, or that certain macro-level phenomena are difficult to derive from micro-level explanations. The actual reason the term has become so prominent over the past two decades is that it is repeatedly invoked in debates about mind, consciousness, meaning, intentionality, representation, content, normativity, teleology, and phenomenal experience.
That is where the real pressure lies. So while a historically careful and philosophically modest account of emergence is valuable, and Humphreys’s book offers an excellent one, it does not fully address the use of emergence that I find most puzzling today. The word is rarely deployed merely to describe a complicated relation between domains. It is increasingly being asked to do heavy explanatory work in some of the hardest problems in philosophy of mind and cognitive science.
For that reason, I am ultimately interested in emergence only insofar as it claims to illuminate problems such as meaning, intentionality, representationality, content, normativity, teleology, consciousness, and phenomenal experience.
Otherwise, emergence is just one philosophical relation among many.