![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
mtvesselDec 2007
Freedom Evolves - Daniel Dennett - Penguin, 2004
* * * *
I’m afraid that this is going to be a very long and intense review, partly because this is one of the most interesting books I’ve read recently and partly because I’m still not sure what I think of it, so I’ll be working it out as I go along. Sorry about that.
The book sets out to show that the concept of free will is compatible with naturalistic philosophies that eschew supernatural concepts such as souls, gods or Cartesian mind/body duality and rely purely on scientific knowledge. The problem is fundamentally one of determinism - since the physical universe appears to run in a causal fashion (at least above the quantum level), it follows that our behaviour must be determined by external genetic, environmental and historical causes and that therefore our perception of ourselves as free agents able to make independent choices must be an illusion. The unfortunate corollary is that any moral system based on personal responsibility must also be wrong, undermining the legal basis of most modern societies. Dennett argues that on the contrary, free will actually exists, and that this has come about because freedom of action and morality have themselves evolved from simpler beginnings.
Dennett’s first task is to convince us that freedom of action (or at least something that looks like freedom of action) is possible in a deterministic universe. He does this through the concept of “stances”, illustrated using John Conway’s Life Game, where a very simple set of rules is applied at each time step to a grid of “on” or “off” cells. His point is that there are several equally valid ways of looking at the patterns which occur in the Life world. The physical stance simply observes the on/off state of the cells - this is of course completely deterministic as you can always predict at one time step what the state of the cells will be at the next time step (although interestingly not vice versa). However, there are also patterns that persist in a recognisable way, such as “oscillators” that cycle through a number of states and then repeat themselves forever, or “gliders” that move across the life world by reconstructing themselves over a series of time steps in a different position.
When you have patterns which can have a history, you can adopt the design stance where you can start asking questions such as how long a particular pattern will last given the presence of other patterns. Dennett makes the point that with the design stance, the words “inevitable” and “determined” are no longer synonymous. A pattern can, in principle, display avoidance behaviour by detecting the presence of an attacking pattern (this can be done if “gliders” are treated like photons) and, for example, moving out of the way or generating a new pattern that destroys the incoming attacker. The outcome is still predetermined but it is nonetheless meaningful to talk about whether or not a particular pattern avoids a particular fate.
You can go still further, and envision a complex pattern that includes processing capabilities that evaluate and resolve conflicting inputs and triggers appropriate (or inappropriate) responses (this may seem fanciful but Conway has shown that a pattern with the properties of a universal Turing machine can in principle be constructed, albeit in a Life world so large that a computer monitor more than a kilometre across would be needed to display it). This is the intentional stance, where we can start to treat the pattern as an agent that has needs or wants and can make decisions. In a sense it is of course an illusion - the pattern’s “decision” is pre-determined - but it is still a valid way of describing higher order regularities in the Life world.
Dennett next takes on the quantum mind theorists such as Roger Penrose and Robert Kane, who argue that a gap must exist in the deterministic processes of decision making within an individual that enables them to take ultimate responsibility for their actions, and that this could consist of an (as yet unknown) amplification mechanism of quantum indeterminacy in the brain. Dennett quite sensibly points out that adding what is in effect a random number generator to our faculty of judgement doesn’t square with what most of us consider moral responsibility to be. Most decision-making has in practice got to be deterministic (based on circumstances - Luther’s “I can do no other”), leaving only rare and somewhat arbitrary windows of opportunity (self-forming actions, or SFAs) where indeterminism could act. And if an undetermined event is a necessary precursor of the SFA (in order to break the deterministic chain), how can the individual be morally responsible for the SFA (and therefore its consequences), when it is based on an event over which, by definition, they have no control? In any case there is no practical difference between an SFA generated by a truly undetermined event and one generated by a pseudo-random (determined but with very complex causes) event.
We now embark on our evolutionary journey towards (so Dennett claims) free will. The first step is relatively easy. There is a clear pressure in a complex environment for organisms to evolve from action-reaction machines, which implement rules of the form “if you encounter condition C, do action A”, to choice machines which consider multiple possible actions using the form “if you encounter condition C, doing action A will result in outcome Z with probability p”, because, as Karl Popper put it, it allows your hypotheses to die in your stead. This evolution of agents which are choice machines achieves a basic level of freedom - a bird can fly where it wants - but clearly isn’t human free will. To get to that, Dennett argues that we need to consider human cultural evolution (via memes) and in particular the development of cooperative and altruistic patterns of behaviour.
This leads into the extremely familiar territory of the Prisoner’s Dilemma and other forms of game theory, but Dennett makes some interesting points. He notes that as the models become more and more “real world”, more and more altruistic-like behaviour appears. Consider, for example, a population of co-operators in which “free-loaders” randomly appear. Obviously the free-loaders eventually take over. Except that they don’t, because the free-loading lifestyle depends on the existence of co-operators. What you end up with is a predator-prey equilibrium. If you then introduce the concept of geographical proximity, where agents with similar addresses are more likely to interact, neighbourhoods of co-operators arise because you are more likely to encounter your own kind, and develop still further if you now give your agents some element of “choice” in the other agents with which they interact, for example ostracism of defectors.
So simple models suggest that social and cooperative behaviour can arise from evolutionary “arms races”, just as genes evolved to cooperate in genomes and cells evolved to cooperate with other cells in multicellular organisms, but we are still not at free will. For a start they do not explain why we can resist the temptation to do the wrong thing when the circumstances arise. Dennett suggests that the development of self-restraint could be an evolutionary answer to the problem of reward curves (our tendency to choose an immediate but lesser reward over than a future but greater one). Applied to our relationships with other people, one can see how lawful behaviour - moral memes, if you like - could arise.
Dennett next tackles perhaps the most serious scientific objection to the idea of free will, which is the notorious experiment of the neurophysiologist Benjamin Libet. He found that when a subject was asked to make a voluntary muscular movement (a wrist flick) at any time they chose, the electrical activity in the brain that leads to the movement (the readiness potential) started some 300 to 500 milliseconds before the subject reported that they made the decision. This implies that the decision to flick your wrist (or fire a gun, or tell someone you love them) is actually made by subconscious processes, with your conscious self only having at best the power of vetoing an action once it has already begun - in short, free won’t rather than free will. Dennett suggests some other interpretations. For example, the time when the decision was made was ascertained by the subject reporting the position of a dot of light that was rapidly rotating about a clock face. Perception of the dot is itself a neural event, so perhaps your brain simply misjudges the simultaneity of the two processes. Or perhaps the decision to flick is written in “slow-drying ink” which means that you can act on it immediately but can’t match it with the stimuli from your visual cortex until the ink “dries” after 300 milliseconds. Or the requirement to visually monitor an action that takes time to get started leads to an ingrained habit that current visual signals are always associated with the decision made 300 milliseconds ago rather than the decision being made now (after all, your brain is more likely to be interested in monitoring the action itself rather than the decision to make the action). All of these are plausible, but I think that Dennett is reaching here with his ad hoc hypotheses. Libet’s explanation has the merit of simplicity and should be accepted simply on the basis of Occam’s razor.
And in any case, it doesn’t really matter. Choosing when to flick a wrist is hardly the same as making a considered moral choice. If anything, it’s closer to the “decision” that a tennis player makes to return a fast serve, a simple action-reaction response. Considered thought is different. Yes, the notions that “pop into our heads” are the results of unconscious processing over which we have no control. But as long as there is an interplay between conscious processes and unconscious ones - as long as we can consciously suppress undesirable or unsuitable thoughts (and veto any bodily actions that they may have triggered), and conversely consciously “encourage” certain notions to appear (like remembering someone’s name), then in my book we still have the potential for free will.
Interplay is also at the heart of Dennett’s final explanation of free will, although at the cultural level rather than the personal. He thinks that our ability for self-reflection mediated by speech - for asking why you are doing what you are doing, both of ourselves and of others - opens up a new evolutionary “niche” for directed selection of the moral memes that inhabit our brains and hence for new forms of self-control. This allows the iceberg of ethics finally to drift into view. We can ask which ethical principles (memes) work and which don’t, and discard the latter.
This bootstrapping approach to the evolution of humans as free beings - that changes in our mental or cultural environment caused by competition between replicating entities (genes or memes) complicate those environments and thereby creates new niches where “higher” forms of evolution can lead to more sophisticated and freer agents - does seem to me to have some merit, although I don’t think that Dennett makes a good job of explaining it in this book. There are a lot of interesting thoughts on a huge range of subjects here, but what with the various digressions into quantum mind speculations, game theory and neuroscience, and the ripostes to critics, it is hard for a non-specialist to follow the thread of the philosophical argument. I have a nasty feeling that Dennett is bludgeoning readers into agreeing with him by bombarding them with such a stream of ideas that their critical faculty gets punch-drunk and they fail to spot the weak links in his chain of reasoning (not intentionally, I think - I’ve been guilty of the same thing). For me, the weakest point in his argument is the use of memes, which are so ill-defined and unfalsifiable that they can be used in just-so stories to “explain” just about any cultural phenomenon and so, of course, explain nothing at all.
There is also the question of whether his evolutionary theory explains not the origin of free will, but the origin of the illusion of free will. After all, the fact remains that the freedom of action of patterns in the Life world is in at least one sense illusory - seen from the physical stance, there is only one pre-determined outcome of any “choice” that the pattern makes - and by analogy this will also be true for any deterministic decision-making apparatus in the real world. Dennett addresses this objection at some length - it is, he claims, a common misunderstanding of his ideas - using an imaginary opponent called Konrad. His ripostes are convincing, but ultimately boil down to the view that a deterministic mechanism using pseudo-random inputs can produce just as good an approximation of “independent” decision-making as a non-deterministic one. In other words, depending on the stance that you choose to adopt you could say that our capacity for free will is an illusion, but if so it’s a jolly good one and that’s reason enough to assign moral weight to our decisions.
Nonetheless, my instinct is that he is on the right track, and that the explanation of the mystery of free will lurks somewhere in the soup of ideas that he presents. But I would like to see someone present it in a clearer form.
Freedom Evolves - Daniel Dennett - Penguin, 2004
* * * *
I’m afraid that this is going to be a very long and intense review, partly because this is one of the most interesting books I’ve read recently and partly because I’m still not sure what I think of it, so I’ll be working it out as I go along. Sorry about that.
The book sets out to show that the concept of free will is compatible with naturalistic philosophies that eschew supernatural concepts such as souls, gods or Cartesian mind/body duality and rely purely on scientific knowledge. The problem is fundamentally one of determinism - since the physical universe appears to run in a causal fashion (at least above the quantum level), it follows that our behaviour must be determined by external genetic, environmental and historical causes and that therefore our perception of ourselves as free agents able to make independent choices must be an illusion. The unfortunate corollary is that any moral system based on personal responsibility must also be wrong, undermining the legal basis of most modern societies. Dennett argues that on the contrary, free will actually exists, and that this has come about because freedom of action and morality have themselves evolved from simpler beginnings.
Dennett’s first task is to convince us that freedom of action (or at least something that looks like freedom of action) is possible in a deterministic universe. He does this through the concept of “stances”, illustrated using John Conway’s Life Game, where a very simple set of rules is applied at each time step to a grid of “on” or “off” cells. His point is that there are several equally valid ways of looking at the patterns which occur in the Life world. The physical stance simply observes the on/off state of the cells - this is of course completely deterministic as you can always predict at one time step what the state of the cells will be at the next time step (although interestingly not vice versa). However, there are also patterns that persist in a recognisable way, such as “oscillators” that cycle through a number of states and then repeat themselves forever, or “gliders” that move across the life world by reconstructing themselves over a series of time steps in a different position.
When you have patterns which can have a history, you can adopt the design stance where you can start asking questions such as how long a particular pattern will last given the presence of other patterns. Dennett makes the point that with the design stance, the words “inevitable” and “determined” are no longer synonymous. A pattern can, in principle, display avoidance behaviour by detecting the presence of an attacking pattern (this can be done if “gliders” are treated like photons) and, for example, moving out of the way or generating a new pattern that destroys the incoming attacker. The outcome is still predetermined but it is nonetheless meaningful to talk about whether or not a particular pattern avoids a particular fate.
You can go still further, and envision a complex pattern that includes processing capabilities that evaluate and resolve conflicting inputs and triggers appropriate (or inappropriate) responses (this may seem fanciful but Conway has shown that a pattern with the properties of a universal Turing machine can in principle be constructed, albeit in a Life world so large that a computer monitor more than a kilometre across would be needed to display it). This is the intentional stance, where we can start to treat the pattern as an agent that has needs or wants and can make decisions. In a sense it is of course an illusion - the pattern’s “decision” is pre-determined - but it is still a valid way of describing higher order regularities in the Life world.
Dennett next takes on the quantum mind theorists such as Roger Penrose and Robert Kane, who argue that a gap must exist in the deterministic processes of decision making within an individual that enables them to take ultimate responsibility for their actions, and that this could consist of an (as yet unknown) amplification mechanism of quantum indeterminacy in the brain. Dennett quite sensibly points out that adding what is in effect a random number generator to our faculty of judgement doesn’t square with what most of us consider moral responsibility to be. Most decision-making has in practice got to be deterministic (based on circumstances - Luther’s “I can do no other”), leaving only rare and somewhat arbitrary windows of opportunity (self-forming actions, or SFAs) where indeterminism could act. And if an undetermined event is a necessary precursor of the SFA (in order to break the deterministic chain), how can the individual be morally responsible for the SFA (and therefore its consequences), when it is based on an event over which, by definition, they have no control? In any case there is no practical difference between an SFA generated by a truly undetermined event and one generated by a pseudo-random (determined but with very complex causes) event.
We now embark on our evolutionary journey towards (so Dennett claims) free will. The first step is relatively easy. There is a clear pressure in a complex environment for organisms to evolve from action-reaction machines, which implement rules of the form “if you encounter condition C, do action A”, to choice machines which consider multiple possible actions using the form “if you encounter condition C, doing action A will result in outcome Z with probability p”, because, as Karl Popper put it, it allows your hypotheses to die in your stead. This evolution of agents which are choice machines achieves a basic level of freedom - a bird can fly where it wants - but clearly isn’t human free will. To get to that, Dennett argues that we need to consider human cultural evolution (via memes) and in particular the development of cooperative and altruistic patterns of behaviour.
This leads into the extremely familiar territory of the Prisoner’s Dilemma and other forms of game theory, but Dennett makes some interesting points. He notes that as the models become more and more “real world”, more and more altruistic-like behaviour appears. Consider, for example, a population of co-operators in which “free-loaders” randomly appear. Obviously the free-loaders eventually take over. Except that they don’t, because the free-loading lifestyle depends on the existence of co-operators. What you end up with is a predator-prey equilibrium. If you then introduce the concept of geographical proximity, where agents with similar addresses are more likely to interact, neighbourhoods of co-operators arise because you are more likely to encounter your own kind, and develop still further if you now give your agents some element of “choice” in the other agents with which they interact, for example ostracism of defectors.
So simple models suggest that social and cooperative behaviour can arise from evolutionary “arms races”, just as genes evolved to cooperate in genomes and cells evolved to cooperate with other cells in multicellular organisms, but we are still not at free will. For a start they do not explain why we can resist the temptation to do the wrong thing when the circumstances arise. Dennett suggests that the development of self-restraint could be an evolutionary answer to the problem of reward curves (our tendency to choose an immediate but lesser reward over than a future but greater one). Applied to our relationships with other people, one can see how lawful behaviour - moral memes, if you like - could arise.
Dennett next tackles perhaps the most serious scientific objection to the idea of free will, which is the notorious experiment of the neurophysiologist Benjamin Libet. He found that when a subject was asked to make a voluntary muscular movement (a wrist flick) at any time they chose, the electrical activity in the brain that leads to the movement (the readiness potential) started some 300 to 500 milliseconds before the subject reported that they made the decision. This implies that the decision to flick your wrist (or fire a gun, or tell someone you love them) is actually made by subconscious processes, with your conscious self only having at best the power of vetoing an action once it has already begun - in short, free won’t rather than free will. Dennett suggests some other interpretations. For example, the time when the decision was made was ascertained by the subject reporting the position of a dot of light that was rapidly rotating about a clock face. Perception of the dot is itself a neural event, so perhaps your brain simply misjudges the simultaneity of the two processes. Or perhaps the decision to flick is written in “slow-drying ink” which means that you can act on it immediately but can’t match it with the stimuli from your visual cortex until the ink “dries” after 300 milliseconds. Or the requirement to visually monitor an action that takes time to get started leads to an ingrained habit that current visual signals are always associated with the decision made 300 milliseconds ago rather than the decision being made now (after all, your brain is more likely to be interested in monitoring the action itself rather than the decision to make the action). All of these are plausible, but I think that Dennett is reaching here with his ad hoc hypotheses. Libet’s explanation has the merit of simplicity and should be accepted simply on the basis of Occam’s razor.
And in any case, it doesn’t really matter. Choosing when to flick a wrist is hardly the same as making a considered moral choice. If anything, it’s closer to the “decision” that a tennis player makes to return a fast serve, a simple action-reaction response. Considered thought is different. Yes, the notions that “pop into our heads” are the results of unconscious processing over which we have no control. But as long as there is an interplay between conscious processes and unconscious ones - as long as we can consciously suppress undesirable or unsuitable thoughts (and veto any bodily actions that they may have triggered), and conversely consciously “encourage” certain notions to appear (like remembering someone’s name), then in my book we still have the potential for free will.
Interplay is also at the heart of Dennett’s final explanation of free will, although at the cultural level rather than the personal. He thinks that our ability for self-reflection mediated by speech - for asking why you are doing what you are doing, both of ourselves and of others - opens up a new evolutionary “niche” for directed selection of the moral memes that inhabit our brains and hence for new forms of self-control. This allows the iceberg of ethics finally to drift into view. We can ask which ethical principles (memes) work and which don’t, and discard the latter.
This bootstrapping approach to the evolution of humans as free beings - that changes in our mental or cultural environment caused by competition between replicating entities (genes or memes) complicate those environments and thereby creates new niches where “higher” forms of evolution can lead to more sophisticated and freer agents - does seem to me to have some merit, although I don’t think that Dennett makes a good job of explaining it in this book. There are a lot of interesting thoughts on a huge range of subjects here, but what with the various digressions into quantum mind speculations, game theory and neuroscience, and the ripostes to critics, it is hard for a non-specialist to follow the thread of the philosophical argument. I have a nasty feeling that Dennett is bludgeoning readers into agreeing with him by bombarding them with such a stream of ideas that their critical faculty gets punch-drunk and they fail to spot the weak links in his chain of reasoning (not intentionally, I think - I’ve been guilty of the same thing). For me, the weakest point in his argument is the use of memes, which are so ill-defined and unfalsifiable that they can be used in just-so stories to “explain” just about any cultural phenomenon and so, of course, explain nothing at all.
There is also the question of whether his evolutionary theory explains not the origin of free will, but the origin of the illusion of free will. After all, the fact remains that the freedom of action of patterns in the Life world is in at least one sense illusory - seen from the physical stance, there is only one pre-determined outcome of any “choice” that the pattern makes - and by analogy this will also be true for any deterministic decision-making apparatus in the real world. Dennett addresses this objection at some length - it is, he claims, a common misunderstanding of his ideas - using an imaginary opponent called Konrad. His ripostes are convincing, but ultimately boil down to the view that a deterministic mechanism using pseudo-random inputs can produce just as good an approximation of “independent” decision-making as a non-deterministic one. In other words, depending on the stance that you choose to adopt you could say that our capacity for free will is an illusion, but if so it’s a jolly good one and that’s reason enough to assign moral weight to our decisions.
Nonetheless, my instinct is that he is on the right track, and that the explanation of the mystery of free will lurks somewhere in the soup of ideas that he presents. But I would like to see someone present it in a clearer form.