Does Everything Have Meaning? | How Machine Learning Theory Helps Understand Psychoanalysis
Does everything have a hidden meaning? What does it mean for something to have a meaning?
Alfred North Whitehead defined symbolic reference as the process through which one aspect of our experience refers to another aspect of our experience. The former is the symbol while the latter is the meaning. For example, if the word “tree” evokes an image of a tree in our mind, the former is a symbol for the latter meaning. But the reverse can also be true: a poet can go in a forest and get inspiration for lyrics. In this way, the image of a tree invokes words such as “tree” in the poet’s mind. Now the image is the symbol and the word is the meaning.
Now, there are a lot of things whose meaning is not obvious at first glance, whom we might be tempted to call meaningless or ‘random’. For many people, dreams seem meaningless because they lack an obvious pattern or structure, they seem completely ‘random’. A better example is a schizophrenic’s speech during a psychotic break. One symptom of psychosis is disorganized speech. A psychotic discourse is definitely ‘nonsense’ in the Deleuzian sense of the word, because we cannot ‘make sense’ of it easily. But does it really lack meaning? Is it true that it does not invoke another aspect of our experience?
In order to answer such a question, we need to add an extra term beyond the symbol/meaning binary. This is because if we want to keep our definition of symbolic reference as ‘the invocation of one aspect of experience from another aspect of experience’, we need to know whose experience are we talking about? Here, Charles Sanders Peirce’s triadic model of sign-object-interpreter can help us. Peirce rejected Saussure’s binary signifier/signified model, instead opting for a triadic model where the interpreter is also taken into account. Here, a sign does not mean something objectively but always means something to someone, and what something means to me may not be the same as what it means to you.
Peirce’s model raises important questions regarding the objectivity of meaning in psychoanalysis. For a psychoanalyst, even seemingly meaningless phenomena such as dreams or slips of the tongue have a meaning. Is there such a thing as a correct interpretation of a subject’s symptom, or do they always mean something to someone in particular? If Peirce is correct, then a dream doesn’t simply have a meaning in of its own, instead that dream means something to someone. When analyzing a dream, we must first ask ourselves “Who speaks?” and “To whom does it speak?”. In the therapeutic setting, the patient’s dream can mean something to the patient and something different to the analyst.
Does classic Freudian psychoanalysis risk imposing the analyst’s meaning upon the analysand through their unresolved counter-transference? Deleuze and Guattari, in ‘A Thousand Plateaus’ argued that this is what Freud did in the Wolf Man case.
Moreover, is it true that in psychoanalysis the analyst must be as impartial as possible and must discover what the symptom means to the analysand and only the analysand? Or should the analyst bring in their own counter-transference to the table in order to create a therapeutic alliance with the analysand where both of them are creating ‘their own world’ with its own meaning and its own language, where both of them give their personal interpretation and significance to the analysand’s symptoms?
It is here that we confront ourselves with one of the limitations of Peirce’s triadic model. Because if all that there is to communication is a sign, an object and its interpreter, then communication would be impossible, since there would be no shared system of signs. If a word means to me something completely different from what it means to you, then how can we communicate? We must introduce a fourth term then: language.
Here I am using the word “language” in a broader sense, referring not only to human language but to any symbolic system of interpretation. A programming language or a system of traffic signs would also be ‘languages’ in the way I am using the word here.
I will define ‘language’ as a system of followable rules for interpreting signs such that multiple interpreters can have an isomorphic relationship to a specific sign. By isomorphic I mean a structure-preserving mapping (a morphism) between two structures of the same type that can be reversed by an inverse mapping. By defining language as an isomorphism, we avoid the risk of falling in either of the two extremes of universalism or relativism. A sign can mean different things to two interpreters speaking the same language, thus avoiding strict universalism or essentialism. But there is still a certain mapping or structure within the relationships between signs, thus avoiding nihilistic relativism where anything can mean anything without any pattern.
Now, we can return to our initial question: does everything have meaning, or are certain acts meaningless? We can answer that a sign can have meaning in one language and be devoid of meaning in another language. For example, the statement “I am debating philosophy” has meaning in English but has no meaning in Romanian. Similarly, the statement “SELECT * FROM Employees” has meaning in the SQL programming language but is meaningless in C++.
The real question now comes: can we extend this concept of languages to seemingly random or nonsensical phenomena? When I ask myself whether dreams have meaning, we can reframe this question as “Is there a language of dreams with followable rules in the same way that spoken languages like English and programming languages like C++ have their own specific syntactic and semantic rules?”. When I ask myself whether a schizophrenic’s words have meaning, I am really asking whether I can discover some sort of ‘language’ of their own, some sort of system of rules that could make me extract useful information from their words. When a schizophrenic speaks incoherently, are they simply speaking in a foreign language that only they understand, or are they speaking in no language at all? Lacan would give the former answers, since he argued that the unconscious is structured like a language, as if the unconscious would speak (through symptoms, dreams, slips of the tongue, delusions) in a foreign language that we humans would need help to understand its rules in the same way we would learn a foreign language like Japanese.
There are still further questions to be asked. If we assume that the unconscious is structured like a language, like Lacan says, is there only one universal language of the unconscious with a universal structure that can be analyzed, or does the unconscious of each individual subject have a different language of its own? Lacan would say that it’s the latter case, while Jung would say that both answers are true. Jung introduced the concept of the collective unconscious, where there are universal patterns (archetypes) that play out in the symptoms and dreams of neurotics and psychotics.
More than this, how can we distinguish when meaning is discovered, created, or both? When someone is speaking to me, I have to interpret their speech, and our common sense assumes that the meaning of their words is discovered and not created. But what about the example of taking or not taking things ‘personally’? When someone is insulting me, I have the freedom to make a choice: do I interpret their insults as an attack on my very person, or do I interpret their words as a signifier for them having a bad day? I have the freedom to think “They really hate my guts” just as I have the freedom to think “They had a bad day and unloaded their emotions on me”. The question that lingers here is whether meaning is discovered or created. In the former case, there would be an objectively correct or true answer to that question (the question of whether they really hate me or whether they just had a bad day). In the latter case, there is no objectively correct answer and it is up to me to create the meaning of their speech. What someone’s words or acts mean to me would be, in this case, an act of (either conscious or unconscious) creation, not an act of discovery.
Another example: let’s say that my romantic partner buys me a gift by surprise. I can choose to interpret their act as a demonstration that they care about me and that they thought about me and wanted to surprise me in order to make me happier. Or I can choose to interpret their act as a demonstration that they think that I’m cheap and that I need their help in order to own material possessions, as if I’m not able to buy those things on my own. Or I can interpret their act as a demonstration that they are trying to ‘bribe’ me because they want to ask me for a favor later. Is one of those three interpretations the objectively correct one, or is it my freedom and only my freedom to choose what something means to me personally?
Viktor Frank, who created logotherapy, also discussed the problem of meaning. He argued that neurotic symptoms such as depression and anxiety can be alleviated by imbuing them with a meaning. He gives the following example in his book “Man’s Search For Meaning”:
“Once, an elderly general practitioner consulted me because of his severe depression. He could not overcome the loss of his wife who had died two years before and whom he had loved above all else. Now, how could I help him? What should I tell him? Well, I refrained from telling him anything but instead confronted him with the question, “What would have happened, Doctor, if you had died first, and your wife would have had to survive you?” “Oh,” he said, “for her this would have been terrible; how she would have suffered!” Whereupon I replied, “You see, Doctor, such a suffering has been spared her, and it was you who have spared her this suffering — to be sure, at the price that now you have to survive and mourn her.” He said no word but shook my hand and calmly left my office. In some way, suffering ceases to be suffering at the moment it finds a meaning, such as the meaning of a sacrifice.”1
Notice the peculiar way Frankl uses the term ‘meaning’ here: suffering is given a meaning, it is imbued with meaning by a subject, in the same way that morphemes have meaning in someone’s words in a certain language. It is almost like the person’s symptom (depression) was speaking in a foreign language to the person, and thus the person could not have a coherent conversation with their depression since they were speaking in different languages. When the person gave meaning to their wife’s death by thinking of it as him sparring her from having to live without him, his melancholia became weaker. It is almost as that act made the person and his symptom start speaking the same language: suddenly, his wife’s death has a meaning he can understand, in the same way that a person who knows English can ‘make sense’ of the words I am writing right now.
The question now is: did the person make the symptom understand his language, did the person learn the symptom’s own language, or did the two create an altogether new language just for them to communicate? Which one of those three interpretations is the correct one? These questions raised by Frankl’s example are not unrelated to the process of psychoanalysis. When a psychoanalyst interprets a dream, are they learning the dream’s language, are they teaching the dream our language, or are they creating an altogether new language just for speaking with the dream in that particular context?
This is a tricky question that I do not have the answer to. But we can rethink our approach to these questions by making an analogy with the way a machine learning model is trained on a set of data. In machine learning theory, a model needs to have a balanced fit on the training data. Underfitting occurs when the model is not complex enough to pick up the underlying patterns in the data, thus having a low accuracy on both the training and the test data. Overfitting occurs when a model is too complex, modelling even the random noise in the data, thus performing well only on the training data but not on the test data. When a model is overfitting, it is almost as if they view every single datapoint as saying something meaningful.
The way a machine learning model learns on its training data is somewhat analogous to the process of meaning discovery and creation within various languages. When a neural network engages in backpropagation, with every new row in the training data it has to make a decision regarding how much it should update each of its weights and biases in light of this new information. If the model would drastically update its model parameters (in this case, the weights and biases) with each and every new piece of data, it risks overfitting the data, and it’s almost as if it viewed every single piece of data as meaningful, as hiding the key to answering what the underlying pattern in the data is.
Perhaps that is also what sometimes occurs in psychoanalysis? When we think that every single element of a dream or every slip of the tongue as having a meaning, are we, in a sense, creating a model which overfits the data, modelling the random noise inside it, thus missing the forest for the trees?
Notice that labelling a certain act as meaningful or meaningless is a decision and not a discovery that can be true or false. In other words, this is a question of ethics and not of epistemology or ontology: whether a certain piece of data is meaningful or not is not a question that has a true or false answer objectively. Instead, we should ask ourselves “Is it useful to consider this piece of data as meaningful?”. After all, this is what machine learning models do as well: with each new piece of data, a neural network has to decide how to adjust its weights and biases in light of this new information. If the neural network decides to not adjust its weights on a certain row from the input data, it’s because it decided that it would not be useful to treat it as meaningful, not that it actually is or is not meaningful in itself.
Because of this, our analogy with machine learning is relevant for both psychoanalytic interpretation as well as for Viktor Frankl’s logotherapy. When Frankl told his patient that his wife’s death meant sparring her the suffering of living without him, this was again not a question of true and false, but a question of it being useful for him to view the situation in that light.
In light of everything I wrote today, perhaps pragmatism is the best philosophy for viewing this problem right now. To the question “Does everything have a hidden meaning?” our answer should then be “We don’t know, but sometimes it’s useful to consider certain things as meaningless”.
NOTES:
1: Viktor Frankl, “Man’s Search For Meaning”, pg. 135
