Are the concepts of fact and value as distinct as we tend to believe?
It is a common understanding that fact and value are distinct categories, mutually exclusive to the extent that philosophy has developed separate fields for the study of each, respectively epistemology and axiology. But is this really so? I shall try below to persuade you that the issue is not as clear-cut as you might believe.
The Twilight of Modernity
The “modern” worldview that there are objectively demonstrable facts about the world which humans are able to ascertain can be dated back to the philosophical investigations of Rene Descartes (“I think therefore I am”) and John Locke in the 17th century, and David Hume and Immanuel Kant in the 18th century in what has become known as the epistemological turn. This led to a dualistic understanding of reality as consisting of an objective (material) realm which is experienced subjectively by human minds. The resulting conceptual separation of mind and body came increasingly to be seen as somewhat artificial, with human experience seen as arising out of a mutual interaction between mind and body. Framing knowledge in this way as depending on experience imbues it with a necessarily subjective element, which tends to undermine the notion of objective truth. Taking this insight to its logical conclusion has resulted in the postmodern worldview which currently dominates humanities and social science departments. In its extreme form this denies objective truth, asserting that what is believed to be true in a society is controlled by those who wield power in that society, in service of their interests. This has led on in turn to the “culture wars,” as opposing sides look to assert their versions of the truth. But that is a story for another day.
Don Trubshaw, in Part 2 of a series of four essays on the subject of “The axiological turn in epistemology” discusses how various philosophers in the 19th and 20th centuries sought to address these issues. In Part 3 of this series he proposes, in partial agreement with the postmodernists, that a resolution of the impasse may be found through an axiological turn, by recognising the primacy of values in our determination of what constitutes true knowledge. He also proposes that knowledge be understood as arising out of a process of engagement of the individual in the various contexts of human existence, e.g. social, cultural, natural and intellectual worlds, each with their own associated values perspectives.
I would look to suggest here that such an axiological turn, while helping address problems arising from the epistemological turn, brings at the same time its own issues. In particular, the loss of an objective basis for knowledge potentially allows power, socio-cultural or political, to fill the vacuum through the imposition of partisan values agendas. We conclude that a thoroughgoing adoption of an engagement paradigm in understanding the origination of knowledge requires a rejection of the idea that either fact-driven or value-driven considerations be considered as primary, but rather that they should be seen as existing in a duality, in other words as two sides of the same coin.
A New View of Mind
The most natural starting point for a discussion of the nature of truth in human society is from an understanding of human cognition, an area where science has made significant forward strides in recent years. In his seminal work The Predictive Mind, Hohwy (2014) set out his view of the mind as a prediction machine, arguing that the strategy by which it makes sense of the world of experience is by minimising prediction error:
A new theory is taking hold in neuroscience. The theory is increasingly being used to interpret and drive experimental and theoretical studies, and it is finding its way into many other domains of research on the mind. It is the theory that the brain is a sophisticated hypothesis-testing mechanism, which is constantly involved in minimizing the error of its predictions of the sensory input it receives from the world. This mechanism is meant to explain perception and action and everything mental in between. It is an attractive theory because powerful theoretical arguments support it. It is also attractive because more and more empirical evidence is beginning to point in its favour. It has enormous unifying power and yet it can explain in detail too. (p. 1)
Since the time of his writing an avalanche of supporting evidence has flowed. The connection between perception and action is important in the theory because not only do our perceptions drive actions, our actions can also facilitate improved perception, as I set out to explain in my recent article on Identity and Worldview. For this reason, the theory is frequently referred to as Active Inference. In terms of the language we introduced above, Active Inference is at its heart a proposal that human cognition needs to be understood in terms of an engagement paradigm.
It should be stressed that, when we talk about “perception” here, we are taking into account all aspects of our conscious experience, including emotional experience. Barrett (2017), in her seminal book on How Emotions Are Made, makes the argument that
Your brain is shaped by the realities of the world that you find yourself in, including the social world made by agreement among people. Your mind is a grand collaboration that you have no awareness of. Through construction, you perceive the world not in any objectively accurate sense but through the lens of your own needs, goals, and prior experience. (p.156)
If it is indeed the case that we perceive the world through such a subjective lens, it might be asked what scope there is for any objective understanding or knowledge? The key thing is that, although we do not share the same lens, we do have much common ground with people around us in terms of needs, goals and interpretations of prior experience. This is a consequence of the fact that, in addition to our sharing a common biology, most of the understanding we accumulate which forms the basis of our lens on (or model of) the world is socially constructed and received by us in curated form, e.g., TV programmes, newspaper articles or Substack blogs like this. Another important factor here is the way in which we validate new information before incorporating it into our model of the world. This often involves taking into consideration what other people think or say, which fact tends to allow a consensus to take shape, which in turn shapes the thinking or interpretation of individuals.
How should we understand the mechanism for such a social consensus to arise out of a model of human cognition which is based on the thought processes of an individual person? The key here is to realise that, despite appearances to the contrary, our beliefs are not unitary. Rather, our model of the world is a Bayesian or probabilistic one, where we entertain multiple beliefs at any one time, but assign a higher probability (or a greater level of trustworthiness) to some than to others. Of course, when asked to make a decision based on our belief, we typically choose the one in which we have the greatest confidence and persuade ourselves (and tell others) that this is our unequivocal, considered opinion.
What happens when evidence is encountered casting doubt on something we previously considered to be true on the basis that it was the best, or most likely, explanation at the time? Our first reaction will usually be a degree of incredulity and, assuming we don’t dismiss the evidence out of hand, we will check whether there might have been some misunderstanding or error. Failing this, we will consider incorporating the new evidence into our model by lowering the probability weighting we associate with our prior belief and increasing the probability weighting associated with an alternative hypothesis offering a better explanation of the new evidence. These are referred to as the posterior probability weightings.
For example, you are out walking on the beach and you see what looks like a penguin swimming in the sea. You immediately dismiss the idea as mistaken as you “know” no penguins live around your area. Then the next day, you hear that some penguins escaped while being transported to a nearby zoo. Naturally, your confidence in your conclusion the previous day is reduced and you assign a significant probability to the new hypothesis of an escapee penguin.
Active Inference as a Two-Way Street
Thus far we have examined how human cognitive strategies work to reach a (probabilistic) understanding of the world we inhabit; in other words how our actions facilitate our inferences. But Active Inference is a two-way street and the inferences we make about the world and what is happening in it drive our actions in turn. There might appear to be two distinct processes at work: actions driving inference and inferences driving action. But this is not how the human brain appears to be configured. Rather it operates as a single system with two-way traffic flow. Based on incoming sensory data, an attempt is made to understand what is happening by consulting our internal model and engaging in actions to test explanatory hypotheses. This generates further data which are used to modify hypotheses and decide which further explorations should be made and so on until we reach the best understanding we believe we can reach. Hohwy (2014) has characterised this as the workings of a hierarchy whose levels engage in “bottom-up and top-down message passing [to] capture otherwise elusive aspects of the phenomenology of perceptual experience.”
So, our model of the world (including of our place in it) is not just used as a basis for understanding what is happening in the world but also to decide what actions we should take to achieve that understanding: decision-making is an intrinsic part of our learning process and vice versa. But in order to choose among multiple possible alternative actions, it is necessary, at least implicitly, to rank them and then choose the best one. This means that there exists a valuing mechanism embedded within our cognitive system which interoperates seamlessly with our sensory perceptions and motor control. More than that, the valuing system itself must be considered part of our internal model of the world and so is constructed through Active Inference just as we proposed the conclusions of our perceptions are.
We have seen therefore how the choice of actions to be taken in pursuit of cognitive tasks requires a values framework for what constitutes a “better” result. The overarching driver in this is to find the most likely explanation of what we are perceiving and why. This is referred to by cognitive scientist Karl Friston as the “low road” to infer the principles of Active Inference.
The Free Energy Principle
However other actions not directly driven by cognitive interest similarly require a values framework to allow optimal decision-making to occur. How does this work? Karl Friston has argued that the more general framework within which Active Inference works is a minimization of what he calls internal free energy, an information theory concept similar to entropy, which needs to be carefully managed by all living systems. This he refers to as the “high road.” As stated by Parr et al. (2022):
Action is quintessentially goal directed and purposive. It starts from a desired outcome or goal …, which is encoded as a prior prediction. Planning proceeds by inferring an action sequence that fulfills this prediction (or equivalently, reduces any prediction error between prior prediction and the current state). (p. 10)
So, what is the driver for actions within this more general framework encompassing potentially all human activity? As we saw above, Barrett (2017) has argued that our cognition is driven by our own needs, goals, and prior experience. In the context of the more general high road framework, the most fundamental human goal or need is to understand the conditions for our well-being and to ensure they are at all times met. A precondition for achieving this is the availability of a model of the world around us and how it impacts (especially in the here and now) on our lives, individually and collectively. But as well as building our model of the world through cognition, as we proposed above, it is necessary to extend that model to incorporate an understanding of what our flourishing in this world looks like. We can then monitor events and identify threats and opportunities which allow us to chart a course which (so far as we are able) optimally steers us towards a life of thriving, for ourselves and those whose well-being our model tells us is an important concern for us.
On the basis that all these additional facets of our internal model are to be understood as driven by the same Active Inference principle, their construction requires actions to be taken for the inference to be possible and the choosing among possible actions in return requires a values framework or hierarchy to be constructed as an intrinsic part of the model. Given that the values framework we construct plays a central role in determining the content of all aspects of our internal model, to the extent that it is an intrinsic part of the model, we realise that the probabilities or weights associated with competing ideas or beliefs which we discussed earlier must be very closely tied in with this values framework.
Conclusion
Indeed it is not difficult to see from a perspective of human cognition that, while the answer to the question which interpretation shall I make of a given sensory experience will involve us choosing the explanation or prediction which is ranked as having the greatest likelihood within our model, the answer to the question what action should I take among various possible options will involve us choosing the one which is ranked as having the greatest likelihood of flourishing. Of course, we make no explicit such deductions or choices, but by virtue of the mechanisms through which our model and associated values hierarchies are constructed, reference to the model implicitly buys into precisely such a mode of reasoning.
Interestingly, our conclusion here that the construction of a values framework is largely about the weights assigned to different values or virtues is echoed in the research of Haidt (2012), who proposed based on his moral foundations theory that the difference in perspective between liberals and conservatives is not so much in the values they espouse as in the weights they assign to them, with liberals in particular citing the embodiment of care (kindness) and the pursuit of equality as the overriding moral imperatives at the expense of essentially all other virtues, such as loyalty/patriotism, sanctity, respect for authority and upholding liberty. I looked at some of the implications of this insight in a recent Substack podcast on The Diversity Deficit in the Moral Sphere.
Although conclusions reached about what gave rise to sensory experiences are referred to as “facts” and conclusions about what gave rise to human actions as “values”, we see that the role these two play in the Active Inference process driving all human cognitive engagement are exactly analogous. As I suggested above, they might by this token be reasonably viewed as two sides of a single coin.
References
Barrett, L. F. (2017). How Emotions Are Made: The Secret Life of the Brain. Macmillan
Haidt, J. (2012). The righteous mind: Why good people are divided by politics and religion (1st ed.). Vintage Books.
Hohwy, J. (2014). The Predictive Mind. Oxford University Press.
Parr, T., Pezzulo, G., & Friston, K. J. (2022). Active Inference: The Free Energy Principle in Mind, Brain, and Behavior (1st ed.). MIT Press.
