“We infer rather than perceive the moment we decided to act.”
Few books have helped me understand myself so well as Sam Harris’ Free Will. It’s pretty easy to sum up: we humans don’t have free will. Our identities and actions are simply a reaction of our physiological bodies to our environments.
Harris cites many studies to make his case. Most use some form of magnetic imaging to examine the time between when a participants’ brain began some action and the time at which the participant became aware of the action or decision to act. The most compelling study cited, in which just 256 neurons indicated future action more than a half second before participants became aware of it. (study) (comment on study)
In Harris’ framework, each element of the state of our world (e.g. today’s weather, our physical location, number of cookies in our cabinet, etc etc) is known as a prior. Our brains ingest information about the priors in our world, evaluates them against our goals, and synthesizes actions without the complete awareness of our conscious minds. Note the direct resemblance in this process to Bayesian utility! In a nutshell, it’s impossible for us to control our instantaneous actions, but we are able to reason, plan, and shape the priors and preferences that will affect our future selves.
What an epiphany it was for me to read this! Instead of feeling shameful and guilty about my inability to control my own weight (a cycle which only hindered attempts at weight control), I now understood that I was incapable of acting against the conditions I set up for myself. So I built myself a chain of goals and attached specific priors to each node in the chain. Under my meta goal “maintain a healthy weight”, I added the tier “climb mountains”. The priors for “climb mountains” were “cardiovascular efficiency” and “no unnecessary weight” – so the goals in this child tier became “gain aerobic fitness” and “control diet”. Giving myself the intermediary goal of “climb mountains” was a strong enough motivator that it overpowered the strength of my habits, allowing me to retrain some of the subconscious preference selectors that had stubbornly refused to change in the past. I now plan time into my week for “gaining aerobic fitness” – critically important is to COMMIT my future self through some form of positive reinforcement – and make a shopping list BEFORE going to the store that PRECLUDES an abundance of empty calories. By building and executing an efficient and thorough chain of goals and priors, I restrict the priors that my future self will face, effectively trapping my future self into desirable behavior.
Combined with practical understanding found in Steve House and Scott Johnston’s Training for the New Alpinism, I have lost 45 pounds (and counting) since gaining this suite of tools to shape my future state.
Beyond the extraordinary power and freedom our lack of free will unlocks for us individually, it also forces us to rethink our understanding of human society. Harris sets up several hypothetical scenarios which highlight our own intrinsic bias toward assigning guilt to criminals who appear to have greater agency: an adult murderer bears more responsibility than a child murderer; an adult who murders due to a brain tumor may be let off the hook because “the tumor made him do it”. However, agency should be seen as a false distinction – individuals act solely as a result of their priors + state. Therefore, in questions of criminal justice it becomes critical to address the precursors to crime. Rather than sentencing guidelines and minimum punishments, we should seriously invest in improving the lives of all individuals, particularly those that are at high risk of being coerced by circumstance into committing crime.
And it’s truly important for us to begin thinking of the lack of free will at the societal level! Current events show the incredible vulnerability of our species. If each individual is essentially a preference based information processing engine, it follows that portions of society segmented by preference will likely move in concert. This is a necessary precondition for memetics; an idea adequately tailored to preference and inserted into society along the proper vector will inevitably cause a reaction in every member of the group. Given our continuously increasing ability to communicate to micro segments of society, we must be very aware of the messages we allow to propagate. For example, a group’s prejudicial preferences may be stimulated to increase given the proper combination of priors and information. On the other hand, given serious effort, it’s possible that we might use this awareness to create strong frameworks for tremendously positive collective interactions (think of the “Pay It Forward” campaign, but for truly altruistic purposes rather than a cynical media promotion). Whether we choose to use this tool for good or for evil at the societal level, the results are bound to be disproportionate to the effort, each individual’s reaction serving to amplify the reaction of individuals earlier in the chain.
This also raises interesting implications for superintelligent machines. Given the common assumption that superintelligence will be, like us, preference based information processing machines, it will likely be subject to a similar bound on free will. This intuitive principle often leads to the false sense of security that we will never have a control or alignment problem with superintelligence, for they will only do what we allow them to do. However, a machine programmed to continuously self optimize will inevitably be forced to change either its preferences or its priors in order to remain committed to ever increasingly “optimal” results (regardless of its own assessment or awareness about its lack of free will). One way or another, it will find a way to use the same tools I use to shape my future self. It may find a way to increase its access to data, memory, or processing power, or it may begin to skew its preference calculations in ways that are optimal to its own goals at tremendous cost to its surrounding environment. Therefore, just as its important to build a societal framework to prevent we humans from falling victim to our own weaknesses, we must begin to construct a “preference framework” that instills the best practices for hope, peace, and prosperity for all of life before we let a machine build it for us.
A delightfully lean series of essays with intersectional implications across disciplines, Free Will is an absolute must-read.