Featured Post

Table of Contents

Click the on "Table of Contents" link above to navigate the thoughts of KLK. - Click on links below to access whole threads or...

Saturday, November 28, 2015

Free Will #6 – Free Will Experiments – Soon et al., 2013

Note:  subsequent to my writing of this post, an excellent scientific paper has been published that makes many of the same arguments, but does it much better.  The paper is by Brass et al., 2019, and I highly recommend reading it if you are interested in this topic.  You can find it [here].  They also present a very interesting model of decision-making that I'm sure will be the topic of some future posts!

         As I discussed in my previous post on this topic, it seems that free will should be observable in the neuron and therefore should be testable.  I’m not the only one to think that way, and in the last few years a few researchers have attempted to design experiments that would test this concept.  I find the experiments to be very interesting and creative, but, unfortunately they fall short of the goal.  I think it is worth discussing here what some of the difficulties are in some detail.  This post is going to have to be longer than most.

          The most commonly referenced paper on this topic is Libet et al. from 1983, but I would rather review in detail a more recent paper:  Soon et al., 2013 “Predicting free choices for abstract intentions.”  It’s an interesting experiment and illustrates many of the basic principles and basic difficulties with this type of experiment.

          The basic principle that is put forth in these papers is that there is activity in the brain that precedes the conscious awareness of “free will” decisions that we make.  The implication is that freely made decisions are not really free – rather, they are the result of earlier unconscious processes in the brain that determined the decision.  Ultimately, what these researchers would like to say is that our decisions are made unconsciously and therefore determined by other brain processes, thus supporting the idea that free will is only an illusion.[1]  Of course the popular press loves that, and is quick to jump to conclusions about the meaning of the research.  Well, I don’t think it means what they think it means!

          First, I’m going to have to explain the methods used in the paper in a bit of detail so that my comments make sense.  The paper itself is freely available at Pubmed, and if you have interest in this topic, it is worth reading the original paper.

          The attempt in this research was to try to get subjects to make a “freely made decision” involving abstract thought.  Previous published experiments of this nature usually involved a task that required physical movement.  Movements involve a certain amount of pre-planning in the brain and the investigators wanted to separate the abstract decision from the brain activity involved in movement.

          In the experiment, subjects were shown a screen with a few numbers on it and one letter in the center.  The screen changed once every second to a new set of numbers and a new letter.  As the screens would move past, the subject was to randomly decide at some point that, in the subsequent screen, they were going to either add or subtract in the screen that followed.  They had to remember the letter that was on the screen when they consciously decided they were going to add or subtract.  The next screen gave them the first number (a single digit) and then the third screen gave them the second number along with four answers and they had to pick the correct one that corresponded to either the addition or subtraction problem.  Then, a fourth screen appeared that had four letters to choose from and they had to pick the letter that was on the screen when they decided to either add or subtract.  That completed one test, and then the process started over again.  One annoying thing for me is that I could not find anywhere in the paper that described how many trials each subject did.  Knowing the number of samples is critical in trying to understand the detailed results and the statistics presented.  Maybe I missed it somewhere in the paper – otherwise it seems like a pretty glaring omission that peer-review would have caught.

          There was a screening process for subjects to make sure they met certain criteria (more on that later) and then 22 subjects completed the real test.  They were positioned in an MRI machine that could scan their brain activity every two seconds.  I won’t go into the details, but there is a method of brain scanning developed a few years ago that shows the areas of the brain with the highest relative activity.  So, the researchers searched across the entire brain to see if there were regions of the brain that had activity related to the decision to add/subtract and the timing of that decision.  Then, since they knew when the subjects were conscious of their decision (based on the design of the test), they looked at brain activity up to eight seconds prior to the conscious decision point. 

          It’s a fairly complicated procedure to determine whether activity is correlated with the decision made, but in the end they found two areas of the brain that had activity that predicted the impending decision with around 59% accuracy.  Practically, this means that you could query this area of the brain and ask “is the person going to make a decision to add/subtract?” and the activity of the area could be analyzed to predict yes or no with respect to an upcoming decision in 59% of the cases.  Note that if you knew nothing you should be able to predict decision or no decision with 50% accuracy.  Nonetheless, 59% is statistically significant, meaning that it is highly probable that the predicted activity really was correlated with the decision and didn’t just happen by chance.  This predictive activity occurred 2-4 seconds before the person was conscious of their decision.  The implication is that some area of your brain knew of the decision before “you” did, thus showing that the decision was not really “free” but was determined by unconscious brain activity.

          I will quote their conclusion here just for completeness before diving into my review of this paper:  “To summarize, we directly investigated the formation of spontaneous abstract intentions and showed that the brain may already start preparing for a voluntary action up to a few seconds before the decision enters into conscious awareness. Importantly, these results cannot be explained by motor preparation or general attentional mechanisms. We found that frontopolar and precuneus/ posterior cingulate encoded the content of the upcoming decision, but not the timing. In contrast, the pre-supplementary motor area predicted the timing of the decision, but not the content.

          First, I’m going to make three general statements and explain their impact on the understanding of the results of this experiment and similar experiments, then I have a few specific comments that would be more like comments I would have made if I had been a reviewer of this paper.

          1.  It’s not clear to me that “consciousness” and “will” are the same things.  The key assumption in these types of experiments is that we are conscious of the will’s decision at the moment it is made.  I’m not certain that such an assumption is valid.  I don’t have any data to prove the correlation one way or the other, but since free will and consciousness are both somewhat nebulous and hard to define, it seems like a bit of a leap to think that they must be perfectly correlated in time.  Isn’t it possible that your will begins to act on your brain before you are conscious of it?
          About the only example I can bring up here is to consider the common phrase “go with your gut.”  What do people mean when they say that?  Physiologically it doesn’t make a lot of sense, yet we all know what is meant by that phrase because we experience it.  Sometimes we know what we need to do right away – almost unconsciously it seems – and yet sometimes we are hesitant.  So people say “go with your gut” because there are times where it almost seems like working through a conscious calculated decision is just not warranted or can even lead us to make the wrong decision.  To me, this indicates that sometimes our will is active before we are conscious of it.  Of course, for those of you who are determinists, you’ll say that this common phrase is as close as we usually come to realizing that every decision we make is pre-determined by our ongoing brain activity.  Possibly.  The point is – who knows?  And until we know, I find it hard to draw firm conclusions from these types of experiments about free will.  To be fair, I don’t think the authors of these papers make this direct connection in their papers – I think it is usually the things written later about the article that make this claim because such a claim is a bit more sensational.
          Along with this issue is that it seems to me that it is particularly difficult to pinpoint when you are conscious of a decision.  Subjects in these studies have to have some way of marking the time point when they are conscious they are making a decision.  I find that conceptually difficult.  How do you really know?  How accurate can you really be?  Seems to me that an error of a few seconds is certainly possible.

          2.  The tasks in these experiments are basically random, not necessarily decisions of the will.  To be honest, I’m not sure how else they could do these experiments.  But I would say that the influence of free will in the human mind and brain is fundamentally different than making random decisions.  I tried to illustrate that a bit with my entry on the “Turing Test with Numbers”.  The difference between a random number series and a series determined by the “will” is hard to identify.  Further, how do human beings make random decisions anyway?  My hypothesis is that the process for picking random numbers is different than the normal process for making decisions.  I would not expect our free will to be involved in the selection of truly “random” decisions or actions.  I kind of figure that we must pick random numbers in a manner similar to the way computers pick random numbers – they have to find some kind of electronically noisy process that has apparent randomness in it.  If so, then we would use some neurons who are tuned to fire or not fire based on small differences in the random fluctuation of ions at the nerve membrane.  I don’t know if this is true or not, but I really don’t see why we need our free will to make a random decision.  If our random decisions are really pre-determined by various brain circuits, that would not bother me at all.  And it has no impact on what I consider free will.

          3.  The will is a weak force.  When determinists argue against free will being a separate influence on the brain, they seem to imply that the will takes over all of the brain’s activity and is the overwhelming influence over what happens.  I don’t see it that way.  I think most of what we do, and most of what the brain does, is essentially composed of physical, material reactions and responses to various sensory inputs.  Essentially, I would say that we are primarily acting on complex reflexes that are fine-tuned by the will.  I think we can go through our day pretty mindlessly doing the things we need to do without involving the will much.  In many cases, I would say that the will’s influence is very subtle.  Frequently I think our will is entirely masked by these reflexes.  Why do we do things that we say we don’t want to do?  We intend to do good things, but then we make bad decisions.  Sometimes in these cases I would say we are driven by our habits and passions, and the will is completely pushed aside.[2] 
          The implication of the will being a weak force is that when we try to experimentally identify it, it is very difficult.  We end up trying to measure a very small signal riding on top of a huge signal, and that is very hard to do.  We are trying to pick out subtle neural activity in the midst of huge networks of coordinated neurons being driving by prior neural inputs.  Fundamentally, this is why I think an experiment demonstrating whether or not we really have free will is extremely difficult to conduct.
          Related to this issue is the method used to identify brain activity.  fMRI is used to identify this activity based on a parameter called BOLD.  This parameter correlates to the use of energy within the neurons.  When neurons are more active, they use more energy than they usually do, and that can be measured.  But this type of measurement requires that there are a large number of neurons in a specific region of the brain that become highly active during any specific time period.  You can’t pick out the activity of a few neurons with this method.  Subtle changes would be lost.  Also, unique activity in broad regions of the brain would not show up here either.  An analogy might be trying to determine where I work by observing the rush hour traffic patterns in Cleveland.  That might predict where a lot of people work, but I happen to work south of downtown (usually).  Others might not even go to work during rush hour times.  The BOLD parameter only picks up major activity and if, as I suggest, the will is a subtle force, it won’t appear in the tests being done using that method.  There are other testing methods used by other investigators, and I may get into those in the future.


Now for some more detailed considerations of this particular paper.

4.  It’s really hard to design an experiment that involves human behavior when you want to give them complete freedom to make a decision.  Experiments need to be well-controlled in order to draw conclusions, while free will decision-making is anything but well-controlled!  So, in this experiment they had to create some artificial constraints on people’s decision-making.  The primary way they did this was to have a screening test in which subjects performed the test (without the MRI) and their spontaneous performance had to meet certain criteria.  This means that the subject population that was studied was a selectively biased population.  Specifically, subjects had to “spontaneously” select about the same amount of addition and subtraction tasks.  The researchers needed reasonable numbers of both types of activity, but how do you guarantee that kind of distribution and still allow people to be “free”?  I don’t think the paper ever said whether the selected subjects continued to pick similar numbers of addition and subtraction tasks when they were in the MRI, but I assume that in general they did.  They do mention that they threw out the data for one subject who never selected a subtraction task in one run.  Obviously, they weren’t really free to choose!
The other thing that I think is even more problematic is that they only selected subjects who averaged at least 10 or more screens before they made a choice.  Thus, subjects who were naturally highly decisive were excluded from this study.  Thus, only subjects who pondered things a bit before “freely deciding” were selected.  Is it possible that “ponderers” have to mull things over before making a decision, thus partially explaining why they had brain activity before they were conscious of their decision?  The bottom line is that they had to evaluate a biased population.

          5.  The best prediction accuracy was about 59%, which is fairly weak and barely above statistical significance, given that by chance the accuracy should be 50%.  But, for an experiment such as this, 59% isn’t terrible.  But there’s some missing information that makes it difficult to judge this predictive ability.  Specifically, the average time for subjects to make a spontaneous decision was 17.8 sec with a standard error of 1.8 sec.  Thus, it seems to me that as time drags on, you can kind of predict about when a decision is going to be made.  Is it really spontaneous in that case?  It seems to me that I made be able to make a 59% prediction just by knowing which screen number the subject was on.  If my brain can predict it, then so can the subject’s!

6.  The activity predicts that a decision is going to be made (fairly poorly), but does not predict what the decision is actually going to be – add or subtract.  The individual is not really free to decide whether to make a decision – the rules of the experiment require that they eventually decide and the rules of practicality essentially require that the decision is made somewhere around 10-20 seconds after the start of each series of screens.  What seems much closer to spontaneous is add or subtract, but that wasn’t predicted by the brain activity.  Again, to be fair, you’d think this would be a pretty subtle activity in the brain and hard to pick out with fMRI methodology.

7.  The subjects had to remember the letter on the screen when they made a decision.  In my opinion (as one who has a bad memory) the anticipation of having to remember something requires brain activity.  In other words, you have to know before you see (or hear) something that it is something you’re going to have to remember.  Otherwise, you immediately forget it (at least that’s how my brain works).  Therefore, I think subjects would have had to do some preparatory brain work prior to seeing the screen where they made a decision, so that their brain was ready to store that letter in memory for about 3 seconds.  So, one hypothesis might be that the regions of the brain they identified were involved in preparatory memory activities, not necessarily in decision-making.  I know they compared the focused activity to general attention mechanisms, but it’s not clear to me that preparation for specific memory is the same as general attention.

8.  One thing this paper does show is that the decision itself and the timing of the decision appeared in separate parts of the brain.  If the will acts at the level of neurons as I propose, this may suggest that the will acts in different areas in the brain and then those areas ultimately need to come together elsewhere in order for a decision to actually be made.  I relate these things to my personal experience (where, in these cases, I’m assuming that I experience what everyone else in the human race experiences).  Sometimes I’ve made up my mind about something, but I’m not ready to actually decide, so nothing happens “until I’m ready.”  The results of this paper might lend some weak credence to that…maybe.


Well, there’s a lot more to be discussed in the various papers in the literature.  The point is, disproving free will and proving determinism is going to be pretty difficult because the experimental design is conceptually very difficult.  I try hard not to be someone who critiques the work of others without offering an acceptable alternative.  I will have to present those ideas in a future entry.  But I hope, at the very least, that this helps you not get carried away by the headlines that say “scientists prove that free will is an illusion.”  They haven’t – and they never will![3]


References

Libet B, Gleason CA, Wright EW, Pearl DK (1983) Time of conscious intention to act in relation to onset of cerebral activity (readiness-potential). The unconscious initiation of a freely voluntary act. Brain 106(Pt 3):623–642.

Soon CS, He AH, Bode S, Haynes JD. Predicting free choices for abstract intentions. Proc Natl Acad Sci U S A. 2013 Apr 9;110(15):6217-22. doi: 10.1073/pnas.1212218110. Epub 2013 Mar 18.




[1] I know:  researchers are supposed to be unbiased on not have anything they “would like to say.”  So far, every researcher I’ve met is a human being and, therefore, biased.  That includes me.
[2] Who “pushes the will aside?”  The will – what else could it be?  It’s not that weak of a force!
[3] Although it’s a little hard for me to prove that something will never happen!

No comments:

Post a Comment