In this chapter, I try to define the word “suffering”. This is not an easy task, as we will quickly see. Defining the term properly requires, to some extent, elucidating the underlying mechanisms creating suffering.
One fundamental point here is that I exclude physical pain from the definition of suffering; I use the word suffering synonymously with mental pain. Nevertheless, I will start the search for a definition of suffering by considering the closely related concept of pain, taken here in the medical sense of physical pain.
The central conclusion of this chapter is that the main definitions of suffering consider it based on either frustration or a threat to the intactness of the person. These two definitions, and especially the definition based on frustration, are the basis of the developments of the rest of this book. From a more abstract viewpoint, I will argue that such suffering can be seen as error signalling, similarly to physical pain.
Let us start by defining pain. Pain has been given a widely accepted consensus definition by the International Association for the Study of Pain (IASP) as
Pain is an unpleasant sensory and emotional experience associated with actual or potential tissue damage or described in terms of such damage.
Surprisingly, while this definition was originally adopted in 1979, it is still used with minimal modifications. It posits damage to any tissue of the person, or any threat of such damage, as the origin of pain. Pain is then defined as an ensuing unpleasant experience. While this definition has been found to be quite useful in a clinical context, deeper theoretical analyses have found various problems.1
One important controversy is whether one should define pain as a subjective experience, or as something that has a more objective existence. The definition above talks about an “experience” which is here interpreted as a conscious, subjective experience: something that only I am aware of, and which you cannot measure in any objective way. As we will discuss in more detail in Chapters 8 and 12, this problem of subjective conscious experience vs. objectively observed phenomena is ubiquitous in neuroscience and psychology.
The problem with talking about such subjective experience in a scientific context is that objective, reproducible measurement is the basis of science. Fortunately, subjective experience can be measured in various indirect ways, such as verbal report. That is, we can ask the patient if there is pain. Yet we will never know for sure what the patient actually feels. In particular, we cannot tell how her experience of pain compares with other people’s experience: Does she feel more pain than some other patient?
This problem in the definition is to some extent alleviated by the reference to tissue damage, which is objectively measurable and reasonably well-defined. Yet, as this definition clearly points out, actual tissue damage is not necessary for pain—since it can be just “potential”—and thus it does not provide a basis for measuring pain or for objectively defining it. (In fact, the definition does not actually say that pain is in any sense proportional to the amount of damage— it is well-known that tissue damage that creates a lot of pain in one person may create little pain in another—which complicates any measurement even more.) Another related problem with the IASP definition above is that it relies heavily on the word “unpleasant”, which is not a very well-defined term, and, again, quite subjective.
One approach to solve these problems is to take an evolutionary approach. To begin with, we could replace “unpleasant experience” in the definition by “experience that has evolved to motivate behaviour, which avoids or minimises tissue damage, or promotes recovery”.2 Here we go towards defining pain using its evolutionary, functional role, while still acknowledging the subjective nature of pain by talking about an “experience”. The downside of such an approach is that it works on a very abstract level, and provides no details on what might cause pain, in contrast to the IASP definition which explicitly points at tissue damage (even if only potential). This definition, in a sense, shifts the burden to understanding the evolutionary goals of certain experiences, which is not easy either. However, one obvious candidate for such an evolutionary goal is minimizing tissue damage and recovering from it, which links this evolutionary approach with the IASP definition. In more general terms, the evolutionary goal could be the maintenance of “homeostasis”, that is, an optimal balance in the physiological condition of the body.3 Such evolutionary logic can be applied on suffering as well, and we will see related argumentation throughout this book.
In contrast to pain, suffering is a rather neglected term in science, and there is nothing like a consensus definition. Intuitively, most people would think suffering also contains an unpleasant feeling or experience as an integral part, while being more abstract and general than physical pain, in particular including more psychological and emotional aspects. A typical dictionary definition is “Feeling of pain or strong stress, either physical or emotional”.4 Like pain, suffering is often considered a subjective experience which cannot be objectively measured.5
One simple and concrete approach to define suffering is to give examples of phenomena related to suffering and possibly producing suffering. A typical list would contain grief, sadness, discomfort, distress, anguish, fear—which is just a random sample, and many different lists can be produced. While this is a good starting point, it does not lead to a solid scientific theory.
Terms such as psychological pain or mental pain are often preferred in neuroscience, and some attempts at definitions of those terms have been made.6 In this line of thinking, suffering is really a generalization of pain. This may not solve the problem of defining suffering, since the burden is then simply shifted to defining pain, but then we can leverage the large literature on pain, in particular the IASP definition just given, as well as any of its critique and improvements.
One approach distinguishes three kinds of pain: physical pain, social pain, and psychological pain.7 An interesting emphasis in this line of research is that all these different kinds of pain are neurally very similar in the sense that the brain areas responsible are the same.8 Here, physical pain is primarily due to physical damage to the body, but it can also be felt when there is a strong anticipation of such physical damage (think about going to a dentist), reminiscent of the IASP definition. In contrast, social pain is an unpleasant feeling due to social exclusion or rejection. Psychological pain is largely the same as what I call mental pain or suffering, and attempt to define here.
Importantly for our purposes, in such an approach, mental or psychological pain is often assumed to be due to reward loss, defined as follows9
[Reward loss is] a negative discrepancy between expected and obtained rewards.
In other words, reward loss happens when you expect a reward but don’t get it, and it leads to mental pain. Reward loss can also be called frustration, although sometimes this term is reserved for the actual suffering caused by reward loss. This provides one important computational viewpoint: reward loss is a function of computations involving expectations, observations of the obtained reward, and their difference.
An alternative approach emphasizes how suffering is related to our person, or self. Psychological or mental pain has been characterized as an aversive state of high self-awareness of inadequacy,10 or a negative appraisal of an inability or deficiency of the self.11 This is analogue to IASP definition of physical pain in the sense that there is “damage”, but on a more mental level, to our image of ourselves as a psychological and social entity.12
A particularly potent and influential idea in this vein is that suffering necessarily involves a threat to, or a loss of, the intactness of the person, as proposed by Cassell:13
Suffering is a state of severe distress induced by the loss of the intactness of person, or by a threat that the person believes will result in the loss of his or her intactness.
This is a natural generalization and abstraction of the IASP definition of pain as related to “tissue damage”. Damage to the intactness of the person includes tissue damage but is something much more general, in particular including damage to one’s self-image. It is of course crucial here to understand what “intactness” means more precisely; Cassell emphasizes the generality of this notion, saying that “suffering may occur in relation to any aspect of personhood”.14
A unifying theory, which combines several different kinds of suffering and pain in a single framework by linking them through the concept of frustration, has been developed by van Hooft.15 He starts from an Aristotelian conception of the human person as having four “parts of the soul”. They range from the lowest level of biological functioning to the emotional/desiring functions and the rational functions, finally reaching the sense of the meaning of existence. In his theory, each of these parts has its own goals, its own form of “fulfillment”, which is again an Aristotelian idea. Suffering is then nothing else than frustration, namely “frustration of the tendency towards fulfillment” of one of the different parts of the soul. In this theory, the lowest level of biological functioning is even below ordinary pain and pleasure, and simply about staying healthy and alive. Ordinary physical pain is the frustration on the emotional/desiring level, where the goal of the organism is to gain pleasure and avoid pain. Frustration of rational (intellectual) function refers to suffering which happens when it is not possible to reach long-term goals that one would usually expect to reach and plan for. Frustration on the highest, “spiritual” level happens when it is impossible to understand why it is me that is sick—in the medical context where van Hooft writes—or life seems meaningless due to the despair and fear which a malady brings with it. This last kind of suffering brings us close to the kind of suffering considered in existential philosophy.16
Closely related definitions can be found in the literature on stress: Lazarus and collaborators define “psychological stress” as “a particular relationship between the person and the environment that is appraised by the person as taxing or exceeding his or her resources and endangering his or her well-being”.17 So, we have to consider the possibility that stress is another kind of suffering, or a mechanism for suffering. However, I don’t take such a view in this book because the classic definition by Hans Selye, “the father of stress”, proposes that “stress is the non-specific response of the body to any demand” (my italics). This is a very general definition, and Selye has explicitly emphasized that positive, happy events can induce stress just as well as negative, threatening ones; think about an athlete engaged in a competition. Based on this definition, it does not seem possible to simply consider stress as one kind of suffering, unless we focus on the negative kind of stress, termed “distress” by Selye.18 The distinction between distress and “pleasant” stress is, unfortunately, not very clear; it has been proposed that it is the unpredictability and uncontrollability of a situation which distinguish the unpleasant distress from other kinds of stress.19 Their connection to suffering will be considered from different viewpoints in this book.
Centuries before any such modern developments, some ancient philosophers already made great progress in understanding suffering. The best expert on the topic may have been the Buddha, and in fact the whole of Buddhist philosophy can be seen as a theory of suffering—especially when considering the original version proposed by the Buddha himself. He gave the following description of suffering:20
Union with what is displeasing is suffering; separation from what is pleasing is suffering; not to get what one wants is suffering.
This is actually not so much a definition of what suffering is, but rather an attempt to describe what the main causes of suffering are.
Stoic philosophers in ancient Greece and Rome had very similar ideas. Epictetus, one of the most famous Stoics, describes mechanisms that lead to suffering as follows:21
[D]esire promises the attainment of that of which you are desirous; and aversion promises the avoiding that to which you are averse. However, he who fails to obtain the object of his desire is disappointed, and he who incurs the object of his aversion wretched.
These are essentially a reformulation of the points given by the Buddha above. We can summarize these philosophical ideas as the following two causes for suffering, each with two variants:22
Then, the definitions by the Buddha and Epictetus can be interpreted in terms of wanting (point 1) and aversion (point 2) only. Point 1 in particular defines the typical case of frustration, related to reward loss considered above. Using the term somewhat liberally, we can also call the suffering in point 2 frustration, since the desire to avoid something is frustrated. Thus, we see that both the ideas of both the Buddha and Epictetus can be simply summarized as saying that suffering comes from frustration.
Now I shall try to recapitulate the ideas above, both ancient and modern, as succinctly as possible. I think we only need to talk about two kinds of suffering, or rather two mechanisms producing suffering, namely:
Furthermore, van Hooft’s theory could be seen as combining these two aspects.
Based on this dichotomy, this book will develop two computational definitions of suffering, one each for these two aspects. Frustration will be specifically dealt with in Chapters 3 and 5, and the threat to the person (or “self”) in Chapter 6. Chapter 6 will also propose how the two aspects can be seen in a unified framework by seeing threats to the person as frustration of certain long-term goals, a bit in the same sense as van Hooft’s theory.
The emphasis in the following chapters is, obviously, on information processing. As already argued in the introduction, my main justification for talking about information-processing is that the framework of information-processing is a practically useful way of describing suffering in the precise sense that it can tell us something about how to reduce suffering. Information-processing is something that we can influence, something we can intervene on, so from a practical viewpoint, it is a very important aspect of suffering to investigate. Focusing on information-processing is also perfectly in line with the current emphasis on cognition in neuroscience and psychology; I see cognition as synonymous with information-processing.
To conclude this chapter, I discuss some computational principles that explain why pain and suffering are so closely related. First, I propose that on a more abstract computational level, both pain and suffering are essentially error signals, messages that something is going wrong from the viewpoint of the goals and rewards of the system. Clearly, frustration signals that something went wrong in terms of not getting what one wants, and a similar case will be made for the threat to the person in Chapter 6. Such error signals are in fact ubiquitous in artificial intelligence, where, in particular, they can be used for learning to choose actions better in view of maximizing rewards. We will see several kinds of error signals in the following chapters, and see how some of them can be interpreted in terms of suffering.
Pain is thus an evolutionarily primitive form of an error signal. Its unique feature is that pain signals are broadcast widely in the information-processing system. This is important in an agent whose computation is distributed into different modules (whether processors or brain regions). For such an agent, it is necessary that any really important signal uses a special pathway that allows it to be broadcast to all, or most of, the modules. The pain signal is indeed broadcast widely to different neural systems, and the signal can change the behaviour of the whole organism in terms of making it stop whatever it is doing and pay close attention to the pain. Furthermore, when an error signal drives the learning of the system, as we will consider in later chapters, it often needs to be observed by several of the modules, and such broadcasting is essential.25
Suffering is largely using the neural systems originally developed for physical pain, as already mentioned. This makes evolutionary sense if we think that computationally more sophisticated forms of error signalling, such as frustration, simply started using the evolutionarily older pain signalling pathway, adapting it for their own purposes. That was practical because the pain system already existed, and served well the purpose of broadcasting error signals to many brain regions. Using the physical pain system for signalling mental pain is thus a useful computational shortcut.26
Yet, merely talking about information-processing, as in a computer, may seem a rather incomplete description of suffering. Why does suffering hurt, if it is merely a signal in an information-processing system? This is in fact exactly the same problem that we encountered with the IASP definition of pain above: Is it a subjective experience, or something more objective and measurable?
The evolutionary rationale just described explains why suffering “hurts” in the same way as physical pain: the physical pain system is hijacked for the purposes of suffering or mental pain. (Perhaps this explains why we talk about mental “pain” in the first place.) The very dichotomy of experience vs. objective measurements is thus exactly the same for pain and suffering, since it is a question of similar experiences and neural pathways. Nevertheless, explaining why physical pain actually subjectively feels like it does in the first place, is an extremely difficult question; it is intimately related to the question of consciousness, which we defer to Chapter 12. We shall rather continue, in the next chapter, by elucidating the computational underpinnings of a particular form of error signal: frustration.