The Science of Experience?

It is amazing how many different definitions there are for science, considering its enormous position in modern society.  And as with many things, inconsistent definitions, or the failure to recognize the use of different definitions, seems responsible for many, many disagreements.  So here is my definition of science.

The minimization of bias from our observations in order to better understand the physical world.

In this framing, science is a tool that we use to inform our understanding of the underlying objective reality so that we can make better predictions about the future.  On some level, all we are ever trying to do is make predictions about the future. Nothing I do can influence the present moment since that moment is gone before I even finish doing the thing in question. I am doing things in any given moment in order to influence the future.  So why should I have any expectations about what my actions will cause in the future?  

Sidestepping the impacts of Hume’s observation about the circularity of inductive inferences, these expectations primarily come from prior experience. In the past, when I have eaten food, my hunger has abated.  Even if I recognize that the expectation that eating in the future will abate future hunger is circular, it nonetheless forms the basis for my actions in order to predictably influence the future.  And maybe we can accommodate Hume by recognizing that our expectations are probabilistic.  The rising of the sun every day in the past is of course no guarantee that it will rise tomorrow, but it does seem to make it very, very, very likely. To the point that we could have a very reasonable expectation that it will rise, even if we can’t be truly certain.

In many ways, this is good enough for an enormous range of actions and expectations about the future. And has been the basis for survival for millions of years.  But what if we do not have sufficient experience with a particular action or phenomenon to have a reasonable expectation?  Or worse, what if we think we know what to expect, but this expectation is actually based on a misunderstanding or misinterpretation of the way the world actually us?  This, by my definition, is bias. Science, then, is the process through which we attempt to remove bias from observations in order to better understand the underlying objective reality. And this allows us to make more accurate predictions about the effect of future actions. 

Bacterial infections are a classic example.  The leading cause of human mortality for all of time until the last half century, the true underlying reality regarding the nature of bacterial infections was poorly understood until around 100 years ago.  Before this time, many thought that the cause of the symptoms and death associated with these infections was excess blood in relation to the other 3 humors (phlegm, black bile, yellow bile). Based on this, bloodletting through the use of leeches, cupping, excision, etc. was considered the standard of care.  There may have been some benefit to those that propagated its use. Certain bacteria require iron reproduce, and blood loss may actually reduce the availability of iron.  This likely propagated the use of this technique through anecdotal evidence (experience).  In this way, we made predictions about the future (recovery from infection) to justify our actions (bloodletting) based on our understanding of the underlying physical world (infections are caused by an imbalance of the 4 humors).  This understanding is based on our observations (of the world itself or of the information provided to us by others).  

With the benefit of hindsight, it is not hard to see the role of bias here. Perhaps somewhat tautological, bias can be defined as anything that causes us to misinterpret our observations about the physical world.  Using the definition above, science is the process through which we identify and minimize bias, thereby allowing our observations to better reveal the true underlying reality.  Recognizing that the history of the scientific method is long and complicated, the high points in this case are the utilization of measurement and experimentation to minimize bias in our observations.

Microscopes allowed for the observation of reality on a new scale, revealing the details of many things that had been previously hidden from us.  While not intentional, our ability to understand the world is inherently limited to information that can be conveyed in a manner that we can sense it.  If we cannot somehow convert information into something that we can see, hear, taste, etc., then how could we possibly observe it?  A microscope bends light in order to magnify images so that we can see objects that our eyes are otherwise unable to see. Presumably these things were always there (and visible/ sensed/known to other species perhaps), but until we had convex lenses to bend light, this information about reality was completely unavailable.  This is a source of bias.  If our predictions about the future are based on our prior observations, then our predictions will only ever be as good as our prior observations. And if our observations do not (or cannot) include the relevant information, then our predictions are likely to be inaccurate.  Microscopes increased the available information in our observations by revealing previously hidden characteristics of the underlying reality.  

But so what?  Just knowing that there are cells and microbes doesn’t improve our ability to treat bacterial infections.  It does, however, allow us to see that the presence of certain bacteria and the presence of certain symptoms (fever, cough, malaise) appear to coincide.  It also allows us to make observations that are reproducible and not dependent on the interpretation of the observer. In other words, the bacteria are visible to virtually anyone who views the same sample through a microscope. It does not require an inference or subjective qualification of the 4 humors that may differ from one individual to the next.  This reproducibility of objective measurement is simply a way to reduce bias in observation.  In other words, if an observation of something is similar across individuals and over time, it is more likely to be a true description of the physical world.

But once the association between the presence of bacteria and certain symptoms is observed, the assumptions that they are the cause of the symptoms and that eliminating the bacteria will reduce the symptoms are still filled with bias.  There is no necessity that either of these is true just based on the observed association.  

Experimentation is the process through which we attempt to remove this form of bias. Leaping forward through the development of antibiotics that work in vitro, we could just give antibiotics to a bunch of people with bacterial infections and see if their symptoms improve. But even if they do, the conclusion that the antibiotics were the cause is still unjustified due to the inherent bias.  Maybe they were all going to recover anyway and the medication had nothing to do with it (similar to how many people currently have an expectation that antibiotics cure viral infections).  The conclusion that the medication caused the improvement is biased, at least in part, on our expectation that it would work.  So how do we identify a causal relationship? In this case, we would reduce the possibility that anything would be related to both the medication administration and the outcome, thereby confounding the association.  We randomly assign people with symptomatic bacterial infections to either receive the medication or not.  Then we compare the outcomes between the groups. In theory, because the groups are formed at random, nothing would be causally related to whether or not they received the medication.  Therefore, any difference between the groups would be attributed to the medication itself.

This is a profoundly oversimplified example, but the point is that science is a tool that we apply in order to reduce the bias that is inherent in our observations of the world so that we can better understand the underlying reality, which allows us to make more accurate predictions about the future.  Under this definition, science is not data, it is not evidence, and the idea of following it seems very confused. It is not one thing with primacy over everything else. It is a tool that can be used to improve our observations.  And it can be used well or it can be used poorly.  And this difference is based on the degree to which it identifies and minimizes sources of bias. 

Importantly, bias can likely never be eliminated. There are unknown sources of bias, and even our hypotheses are due to our prior experiences and expectations.  Science as a process is simply the most effective tool we currently have to attempt to identify and reduce or control for different sources of bias.  When it is done well, it advances our understanding of the true nature of reality (bacteria cause infections), so that we can make better predictions about the future (eliminating bacteria eliminates infections). When done poorly, it can obscure our understanding of the world (vaccinations cause autism) and lead to poor choices based on inappropriate predictions about the future (avoiding vaccines will reduce the risk of autism) But it is important to recognize that we are only aware of the obscured understanding once we apply a process that better reduces the bias relative to the prior science that was utilized.  For example, estimates of the age of the earth ~100 million years based on expectations of cooling times for an originally molten planet were dismissed once radioactivity was discovered and accounted for.  This does not necessarily mean that all of our scientific beliefs are doomed to be disproven as our measurements change, but it does mean that our understanding (and beliefs) is only as good as the degree to which our measurements and experiments reduce bias from observation. And it seems naive to assume that neither of these would improve in the future. 

Under this definition, does science have anything to say about subjective experiences?  If science is a tool to remove bias from our observations about the physical world, what would this even mean in the context of a subjective experience?  By definition the experiences can only be experienced by the subject. But what if there is a physical correlate for all subjective experiences, such that we can “know” what the experience is that someone is having based on some observable physical characteristic?  If there is a physical “signature” (EEG representation of brain activity, eg) for any given experience, then we have effectively “measured” the experience and we can utilize our scientific process to it. Right?

The problem with this is that, by definition, we cannot observe anyone else’s experience.  All we ever get is their description of their own experience, and we correlate that with the physical description of their brain activity.  The bias here, between what they experience and how they describe it (even if done honestly) cannot be avoided.  Because they’re description of it is just based on their prior experience of describing their subjective experiences.  In other words, we have subjective experiences throughout our lives and we are taught by others how to describe them using terms and references so that we can agree on what we are referring to.  When a child has an experience of something soft, it is told that this thing it is touching is “soft” so that the next time they have that experience, they know that what they are feeling can be described as “soft”.  This does not mean that what soft feels like to them is what it feels like to someone else. And there is no way for us to know what it feels like to them. And measurements of physical characteristics of the brain, for example, do nothing to resolve this, since it will only ever tell us what the brain’s response is to feeling something soft, not what it feels like to the person experiencing it.  Even in a deterministic world with an epiphenomenological view of experience, the scientific process would still not be able to bridge this gap.  Science, under this definition, has a lot to say about sensation, but has no role in understanding subjective experience, since the bias that can be introduced between sensation and perception is unavailable to it.

So should there even be scientific efforts to understand consciousness?  I assume the answer to this is at least partly a function of definitions and different definitions of these terms will of course imply a different answer. But if we use the definition of science above and consider consciousness is the whole of subjective experience (what it is like to be something), then I would argue that one has nothing to do with the other.