‘The Way we are Free’ . David R. Weinbaum (Weaver) . ECCO . VUB . 2017
Abstract: ‘It traces the experience of choice to an epistemic gap inherent in mental processes due to them being based on physically realized computational processes. This gap weakens the grasp of determinism and allows for an effective kind of freedom. A new meaning of freedom is explored and shown to resolve the fundamental riddles of free will, ..’. The supposed train of thought from this summary:
- (Physically realized) computational processes underpin mental processes
- These computational processes are deterministic
- These computational processes are not part of people’s cognitive domain: there is an epistemic gap between them
- The epistemic gap between the deterministic computational processes and the cognitive processes weakens the ‘grasp of determinism’ (this must logically imply that the resulting cognitive processes are to some extent based on stochastic processes)
- The weakened grasp leads to an ‘effective kind of freedom’ (but what is an effective kind of freedom? Maybe it is not really freedom but it has the effect of it, a de facto freedom, or the feeling of freedom)?
- We can be free in a particular way (and hence the title).
First off: the concept of an epistemic gap resembles the concept of a moral gap. Is it the same concept?
p 3: ‘This gap, it will be argued, allows for a sense of freedom which is not epiphenomenal,..’ (a kind of a by-product). The issue is of course ‘a sense of freedom’, it must be something that can be perceived by the beholder. The question is whether this is real freedom or a mere sense of freedom, if there is a difference between these.
‘The thesis of determinism about actions is that every action is determined by antecedently sufficient causal conditions. For every action the causal conditions of the action in that context are sufficient to produce that action. Thus, where actions are concerned, nothing could happen differently from the way it does in fact happen. The thesis of free will, sometimes called “libertarianism”, states that some actions, at least, are such that antecedent causal conditions of the action are not causally sufficient to produce the action. Granted that the action did occur, and it did occur for a reason, all the same, the agent could have done something else, given the same antecedents of the action’ [Searle 2001]. In other (my, DPB) words: for all deterministic processes the direction of the causality is dictated by the cause and effect relation. But for choices produced from a state of free will other actions (decisions) are possible, because the causes are not sufficient to produce the action. Causes are typically difficult to deal with in a practical sense because some outcome must be related to its causes. This can only be done after the outcome has occurred. Usually the causes for that outcome are very difficult to identify, because the relation is if and only if. In addition a cause is usually a kind of a scatter of processes within some given contour or pattern, one of which must then ‘take the blame’ as the cause.
‘There is no question that we have experiences of the sort that I have been calling experiences of the gap; that is, we experience our own normal voluntary actions
in such a way that we sense alternative possibilities of actions open to us, and we sense that the psychological antecedents of the action are not sufficient to fix the action. Notice that on this account the problem of free will arises only for consciousness, and it arises only for volitional or active consciousness; it does not arise for perceptual consciousness‘ [Searle 2001]. This means that a choice is made even though the psychological conditions to make ‘the perfect choice’ are not satisfied, information is incomplete or a frivolous choice is made: ‘should I order a pop-soda or chocolate milk?’. ‘The gap is a real psychological phenomenon, but if it is a real phenomenon that makes a difference in the world, it must have a neurobiological correlate’ [Searle 2001]. Our options seem to be equal to us and we can make a choice between various options on a just-so basis (‘god-zegene-de-greep’). Is it therefore not also possible that when people are aware of these limitations they have a greater sense of freedom to make a choice within the parameters known and available to them?
‘It says that psychological processes of rational decision making do not really matter. The entire system is deterministic at the bottom level, and the idea that the top level has an element of freedom is simply a systematic illusion… If hypothesis 1 is true, then every muscle movement as well as every conscious thought, including the conscious experience of the gap, the experience of “free” decision making, is entirely fixed in advance; and the only thing we can say about psychological indeterminism at the higher level is that it gives us a systematic illusion of free will. The thesis is epiphenomenalistic in this respect: there is a feature of our conscious life, rational decision making and trying to carry out the decision, where we experience the gap and we experience the processes as making a causal difference to our behavior, but they do not in fact make any difference. The bodily movements were going to be exactly the same regardless of how these processes occurred‘ [Searle 2001]. The argument above presupposes a connection between determinism and inevitability, although the environment is not mentioned in the quote. This appears to be flawed because there is no such connection. I have discussed (ad-nauseam) in the Essay Free Will Ltd, borrowing amply from Dennett (i.a. Freedom Evolves). The above quote can be summarized as: if local rules are determined then the whole system is determined. Its future must be knowable, its behavior unavoidable and its states and effects inevitable. In that scenario our will is not free, our choices are not serious and the mental processes (computation) are a mere byproduct of deterministic processes. However, consider this argument that is relevant here developed by Dennett:
- In some deterministic worlds avoiders exist that avoid damage
- And so in some deterministic worlds some things are avoided
- What is avoided is avoidable or ‘evitable’ (the opposite of inevitable)
- And so in some deterministic worlds not everything is inevitable
- And so determinism does not imply inevitability
‘Maybe this is how it will turn out, but if so, the hypothesis seems to me to run against everything we know about evolution. It would have the consequence
that the incredibly elaborate, complex, sensitive, and – above all – biologically expensive system of human and animal conscious rational decision making would actually make no difference whatever to the life and survival of the organisms’ [Searle 2001]. But the argument cannot logically be true and as a consequence nothing is wasted so far.
‘In the case that t2>t1, it can be said that a time interval T=t2-t1 is necessary for the causal circumstance C to develop (possibly through a chain of intermediate effects) into E. .. The time interval T needed for the process of producing E is therefore an integral part of the causal circumstance that necessitates the eventual effect E. .. We would like to think about C as an event or a compound set of events and conditions. The time interval T is neither an event nor a condition‘ [p 9-10]. This argument turns out to be a bit of a sideline, but I defend the position that time is not an autonomous parameter, but a derivative from ‘clicks’ of changes in relations with neighboring systems: this quote covers it perfectly: ‘Time intervals are measured by counting events‘ [p 9]. And this argues exactly the opposite: ‘Only if interval T is somehow filled by other events such as the displacement of the hands of a clock, or the cyclic motions of heavenly bodies, it can be said to exist‘ [p 9], because time is the leading parameter and the events such as the moving of the arm of a clock is the product. This appears to be the world explained upside down (the intentions seem right): ‘If these events are also regularly occurring and countable, T can even be measured by counting these regular events. If no event whatsoever can be observed to occur between t1 and t2, how can one possibly tell that there is a temporal difference between them, that any time has passed at all? T becoming part of C should mean therefore that a nonzero number N of events must occur in the course of E being produced from C’ [p. 9]. My argument is that if a number of events lead to the irreversible state E from C then apparently time period T has passed. Else, if nothing irreversible takes place, then no time passes, because time is defined by ‘clicks’ occurring, not the other way around. Note that the footnote 2 on page 9 explains the concept of a ‘click’ between systems in different words.
The concepts of Effective and Neutral T mean a state of a system developing from C to E while conditions from outside the system are injected, and where the system develops to E from its own initial conditions alone. Note that this formulation is different from Weaver’s argument because t is not a term. So Weaver arrives at the right conclusion, namely that this chain of events of Effective T leads to a breakdown of the relation between deterministic rules and predictability [p 10], but apparently for the wrong reasons. Note also that Neutral T is sterile because in practical terms it never occurs. This is probably an argument against the use of the argument of Turing completeness with regards to the modeling of organizations as units of computation: in reality myriad of signals is injected into (and from) a system, not a single algorithm starting from some set of initial conditions, but a rather messy and diffuse environment.
‘Furthermore, though the deterministic relation (of a computational process DPB) is understood as a general lawful relation, in the case of computational processes, the unique instances are the significant ones. Those particular instances, though being generally determined a priori, cannot be known prior to concluding their particular instance of computation. It follows therefore that in the case of computational processes, determinism is in some deep sense unsatisfactory. The knowledge of (C, P) still leaves us in darkness in regards to E during the time interval T while the computation takes place. This interval represents if so an epistemic gap. A gap during which the fact that E is determined by (C, P) does not imply that E is known or can be known, inferred, implied or predicted in the same manner that fire implies the knowledge of smoke even before smoke appears. It can be said if so that within the epistemic gap, E is determined yet actually it is unknown and cannot be known‘ [p 13]. Why is this problematic? The terms are clear, there is no stochastic element, it takes time to compute but the solution is determined prior to the finalization of the computation. Only if the input or the rules changes during the computation, rendering it incomputable or irrelevant. In other words: if the outcome E can be avoided then E is avoidable and the future of the system is not determined.
‘.. , still it is more than plausible that mental states develop in time in correspondence to the computational processes to which they are correlated. In other words, mental processes can be said to be temporally aligned to the neural processes that realize them‘ [p 14]. What does temporally aligned mean? I agree if it means that these processes develop following, or along the same sequence of events. I do not agree if it means that time (as a driver of change) has the same effect on either of the processes, computational (physical) and mental (psychological): time has no effect.
During gap T the status of E is determined by conditions C and P but its specifics remain unknown by anyone during T (suppose it is in my brain then I of all people would be the one to know and I don’t). And at t2, T having passed, any freedom of choice is in retrospect, E now being known. T1 and t2 are in the article defined as the begin state and the end state of some computational system. If t1 is defined as the moment when an external signal is perceived by the system and t2 is defined as the moment at which a response if communicated by the system to Self and to outside, then the epistemic gap is ‘the moral gap’. This phrase refers to the lapsed time between the perception of an input signal and the communicating of the decision to Self and others. The moral comes from the idea that the message was ‘prepared in draft’ and tested against a moral frame of reference before being communicated. The moral gap exists because the human brain needs time to compute and process the input information and formulate an answer. The Self can be seen as the spokesperson, functionally a layer on top of the other functions of the brain and it takes time to make the computation and formulate its communication to Self and to external entities.
After t1 the situation unfolds as: ‘Within the time interval T between t1 and t2, the status of the resulting mental event or action is unknown because, as explained, it is within the epistemic gap. This is true in spite the fact that the determining setup (C, P) is already set at time t1 (ftn 5) , and therefore it can be said that E is already determined at t1. Before time t2, however, there can be no knowledge whether E or its opposite or any other event in <E> would be the actual outcome of the process‘ [p 17]. E is determined but not known. But Weaver counter argues: ‘While in the epistemic gap, the person indeed is going through a change, a computation of a deliberative process is taking place. But as the change unfolds, either E or otherwise can still happen at time t2 and in this sense the outcome is yet to be determined (emphasis by the author). The epistemic gap is a sort of a limbo state where the outcome E of the mental process is both determined (generally) and not determined (particularly) [p 17]. The outcome E is determined but unknown to Self and to God; God knows it is determined, but Self is not aware of this. In this sense it can also be treated as a change of perspective, from the local observer to a distant more objective observer.
During the epistemic gap another signal can be input into the system and set up for computation. The second computation can interrupt the one running during the gap or the first one is paused or they run in parallel. However the case may be, it is possible that E never in fact takes place. While determined by C at t1 not E takes place at t2 but another outcome, namely of another computation that replaced the initial one. If C, E and P are specific for C and started by it then origination is an empty phrase, because now a little tunnel of information processing is started and nothing interferes. If they are not then new external input is required which specifies a C1, and so see the first part of the sentence and a new ‘tunnel’ is opened.
This I find interesting: ‘Moreover, we can claim that the knowledge brought forth by the person at t2 be it a mental state or an action is unique and original. This uniqueness and originality are enough to lend substance to the authorship of the person and therefore to the origination at the core of her choice. Also, at least in some sense, the author carrying out the process can be credited or held responsible to the mental state or action E, him being the agent without whom E could not be brought forth‘ [p 18]. The uniqueness of the computational procedure of an individual makes her the author and she can be held responsible for the outcome. Does this uphold even if it is presupposed that her thoughts, namely computational processes, are guided by memes? Is her interpretation of the embedded ideas and her computation of the rules sufficiently personal to mark them as ‘hers’?
This is the summary of the definition of the freedom argued here: ‘The kind of freedom argued for here is not rooted in .., but rather in the very mundane process of bringing forth the genuine and unique knowledge inherent in E that was not available otherwise. It can be said that in any such act of freedom a person describes and defines herself anew. When making a choice, any choice, a person may become conscious to how the choice defines who he is at the moment it is made. He may become conscious to the fact that the knowledge of the choice irreversibly changed him. Clearly this moment of coming to know one‟s choice is indeed a moment of surprise and wonderment, because it could not be known beforehand what this choice might be. If it was, this wouldn‟t be a moment of choice at all and one could have looked backward and find when the actual choice had been made. At the very moment of coming to know the choice that was made, reflections such as „I could have chosen otherwise‟ are not valid anymore. At that very moment the particular instance of freedom within the gap disappears and responsibility begins. This responsibility reflects the manner by which the person was changed by the choice made‘[pp. 18 -9]. The author claims that it is not a reduced kind of freedom, but a full version, because: ‘First, it is coherent and consistent with the wider understanding we have about the world involving the concept of determinism. Second, it is consistent with our experience of freedom while we are in the process of deliberation. Third, we can now argue that our choices are effective in the world and not epiphenomenal. Furthermore, evolution in general and each person‟s unique experience and wisdom are critical factors in shaping the mental processes of deliberation‘ [p 19]. Another critique could be that this is a strictly personal experience of freedom, perhaps even in a psychological sense. What about physical and social elements, in other words: how would Zeus think about it?
This is why it is called freedom: ‘Freedom of the will in its classic sense is a confusion arising from our deeply ingrained need for control. The classic problem of free will is the problem of whether or not we are inherently able to control a given life situation. Origination in the classic sense is the ultimate control status. The sense of freedom argued here leaves behind the need for control. The meaning of being free has to do with (consciously observing) the unfolding of who we are while being in the gap, the transition from a state of not knowing into a state of knowing, that is. It can be said that it is not the choice being originated by me but rather it is I, through choice, who is being continuously originated as the person that I am. The meaning of such freedom is not centered around control but rather around the novelty and uniqueness as they arise within each and every choice as one‟s truthful expression of being‘ [p 20]. But in this sense there is no control over the situation, and given there is the need to control is relinquished, this fact allows one to be free.
‘An interesting result regarding freedom follows: a person‟s choice is free if and only if she is the first to produce E. This is why it is not an unfamiliar experience that when we are in contact with persons that are slower than us in reading the situation and computing proper responses, we experience an expansion of our freedom and genuineness, while when we are in contact with persons that are faster than us, we experience that our freedom diminishes.
Freedom can then be understood as a dynamic property closely related to computation means and distribution of information. A person cannot expect to be free in the same manner in different situations. When one‟s mental states and actions are often predicted in advance by others who naturally use these predictions while interacting with him, one‟s freedom is diminished to the point where no genuine unfolding of his being is possible at all. The person becomes a subject to a priori determined conditions imposed on him. He will probably experience himself being trapped in a situation that does not allow him any genuine expression. He loses the capacity to originate because somebody or something already knows what will happen. In everyday life, what rescues our freedom is that we are all more or less equally competent in predicting each other‟s future states and actions. Furthermore, the computational procedures that implement our theories of mind are far from accurate or complete. They are more like an elaborate guess work with some probability of producing accurate predictions. Within such circumstances, freedom is still often viable. But this may soon radically change by the advent of neural and cognitive technologies. In fact it is already in a process of a profound change.
In simple terms, the combination of all these factors will make persons much more predictable to others and will have the effect of overall diminishing the number of instances of operating within an epistemic gap and therefore the conditions favorable to personal freedom. The implications on freedom as described here are that in the future people able to augment their mental processes to enjoy higher computing resources and more access to information will become freer than others who enjoy less computing resources and access to information. Persons who will succeed to keep sensitive information regarding their minute to minute life happenings and their mental states secured and private will be freer than those who are not. A future digital divide will be translated into a divide in freedom‘ [pp 23-6].
I too believe that our free will is limited, but for additional and different reasons, namely the doings of memes. I do believe that Weaver has a point with his argument of the experience of freedom in the gap (which I had come to know as the ‘Moral Gap’) and the consequences it can have for our dealings with AI. There my critique would be that the AI are assumed to be exactly the same as people, but with two exceptions: the argument made explicit that 1) they compute much faster than people and the argument 2) left implicit that people experience their unique make-up such that they are confirmed by it as per their every computation; this experience represents their freedom. Now people have a unique experience of freedom that an AI can never attain providing them a ticket to relevance among AI. I’m not sure that if argument 2 is true that argument 1 can be valid also.
I agree with this, also in the sense of the coevalness between individuals and firms. If firms do their homework and such that they prepare their interactions with the associated people, then they will come out better prepared. As a result people will feel small and objectivised. They are capable of computing the outcome before you do hence predicting your future and limiting you perceived possibilities. However, this is still a result of a personal and subjective experience and not an objective fact, namely that the outcome is as they say, not as you say.