'Frankenstein: The Shadow in Artificial Intelligence' is a new, recurring topic in the Unconscious v AI blog series that explores the pros and cons of AI in the workplace, from a uniquely Jungian or analytical psychology perspective.
I was inspired to create Frankenstein: the shadow in Artificial Intelligence blog after I completed the University of Oxford, Saïd Business School's AI Programme and by my psychoanalytic training which emphasizes #metaphor, the definition, and exploration of one thing by reference to the image of another (A Critical Dictionary of Jungian Analysis, Samuels et al, 2013). Metaphor is used extensively in analytical psychology, much like story-tellers and writers do, and in myth, ritual, and religion, as a conscious technique to suggest the subtleties of mystery, or to express something inexpressible in words.
Metaphorical thinking
Analytical psychology sees the psyche as a totality of #conscious and #unconscious processes. The unconscious contains a deep reservoir of images called #archetypes, which are understood by way of a #symbol, defined by Jung as 'the best possible expression of an as yet undisclosed fact'. A symbol is interpreted metaphorically, and by focusing on and remaining as close as possible to the image. Archetypal images are universal and represent the #collectiveunconscious emerging from historic and pre-historic times portraying humans' unlearned behaviour and wisdom. Think of the archetypal images of St George and the Dragon, or Ancient Egyptian Pharaoh Ramses II slaying a captive on the battleground. They are symbolic images that have powerful collective and individual meanings.
The psychological functioning of the totality of the #personality or psyche, known as the #Self, is likened to a #God-image, an image of the totality of one's entire personality or #wholeness, a balance of conscious and unconscious processes, much like homeostasis; a self-regulating process by which a living organism can maintain internal stability while adjusting to changing external conditions. It is a dynamic process that can change internal conditions (unconscious) as required to survive external challenges (consciousness).
The psyche is also seen as self-regulating; constantly regulating itself in an attempt to balance conscious and unconscious processes for the emotional health and well-being of, and to facilitate development and growth in, the individual's psyche.
The relationship between consciousness and the unconscious is a compensatory one. #compensation in psychological terms means balancing, adjusting, or supplementing. The activity of the unconscious balances any tendency towards one-sidedness by consciousness, also known as the #ego or #ego-consciousness, for personality transformation.
Naturally occurring personality traits, for example, extraversion, introversion, and a preference for feeling rather than thinking, of an individual that is repressed, excluded, withdrawn, undeveloped, or inhibited by ego consciousness fall or lapse into unconsciousness and there form a counter-pole to consciousness.
This counter-position strengthens with any increase of emphasis upon the conscious attitude of ego consciousness until it interferes with the activity of ego consciousness itself. The one-sidedness of the personality establishes a balance of disturbed conscious v unconscious equilibrium, also known as a #neurosis, activating not only forgotten personal memories and habitually unconscious and repressed personality traits but also the #shadow which up to then had been intrinsically unknown to the ego.
Neurosis is an expression of a dysfunctional compensatory relationship between conscious and unconscious. In other words, neurotic people experience emotional problems as a result of an imbalance of their conscious and unconscious minds. As their emotional state worsens, the repressed unconscious contents gather a sufficient energy charge to break through into ego consciousness in the form of dreams, spontaneous images, or neurotic symptoms. This is why dreams are a core feature of psychoanalysis; dreams are a gateway to the unconscious and in an analysis can be used for therapeutic relief.
The objective of the compensation is to link, like a bridge, conscious and unconscious minds. However, for the symbol to be effective, the ego-consciousness must acknowledge and understand the symbol both intellectually and emotionally. The unconscious corrects the ego's conscious attitude with the opposite position, or by complementing or affirming it.
The meaning behind the symbol rather than mental treatment relieves the suffering brought about by neurosis. The psyche speaks or reasons in images and the closest rational equivalent is analogy or metaphor. Therefore, the Jungian method of symbolic interpretation #amplification involves using mythic, historical and cultural parallels to clarify the metaphorical content of a symbol. Jung speaks of this as 'the psychological tissue' in which the image is embedded. It involves more than interpretation, it is also a search for a relevant metaphor.
From the metaphor, ego-consciousness may understand a psychological message, while the psyche can re-orient itself by way of an enlarged consciousness through the integration of previously unknown, unconscious facts.
'It's alive!'
My search for a relevant metaphor for the pros and cons of AI is the story of Frankenstein, written by Mary Shelley in 1818. Frankenstein or The Modern Prometheus tells a compelling story about a young scientist, Victor Frankenstein, who creates a *sapient creature, *having or showing great wisdom, sound judgment, or self-awareness, in an unorthodox scientific experiment. The parallels between the story of Frankenstein and the development of AI strike me as 'frighteningly' similar.
The 'Singularity'
You may have heard of concerns about AI. A particular long-term concern is what happens if we reach the ‘singularity’, the hypothesized point at which AI become smarter than people The concern is that AI will use their intelligence to improve themselves, then apply their smarter selves to improving themselves further, and so on (Artificial Intelligence. Wooldridge, Michael. Kindle Edition).
The nightmare scenario is that super-intelligent AI will then be beyond human control and may even pose a threat to our existence. Scenarios like this are the stuff of science fiction, familiar from stories such as Frankenstein, and movies such as Terminator, Megan, and George Lucas' THX1138.
The three laws of robotics proposed by science fiction author Isaac Asimov was an early attempt to develop an ethical framework to curb the potential harm caused by super-intelligent AI.
Law 1
A robot may not injure a human or, through inaction, allow a human to come to harm.
Law 2
Law 3
Unfortunately, Asimov’s Laws require a robot to be able to predict the consequences of all its actions into the distant future. This is too computationally challenging to ever be possible. Scientific opinion is divided about whether the singularity might happen. Even if it did, it is not obvious why the nightmare scenario should occur. There is much uncertainty about the long-term future of AI. For the foreseeable future, though, it is a safe bet that far more people will die from natural stupidity than from artificial intelligence (Artificial Intelligence, Wooldridge, Michael. Kindle Edition).
AI, the new electricity
AI has been described as 'the new electricity', it will transform our personal and work lives, and create huge economic value for businesses and industries that adopt AI to automate tasks and activities. The emergence of AI is a natural progression from the Industrial Revolution, a transition from creating goods by hand to using machines in the period 1760 to 1840. This advancement in working life and practices vastly improved the economies of Western nations. There are three industrial revolutions that transformed modern society. With each of these three advancements - mechanisation and the steam engine, mass production, and electricity and automated production and information technology - the world around us fundamentally changed. Today, we are in the midst of the fourth industrial revolution, digitalization, and AI.
(Artificial Intelligence Programme, University of Oxford, Said Business School)
Geoffrey Hinton, known as the Godfather of AI, just quit Google and says he regrets his life’s work in AI because it can be hard to stop ‘bad actors from using it for bad things’. What does he mean by bad actors using it for bad things? The short answer and an example is algorithmic bias, which this blog series explores, specifically, using the psychoanalytic concept of the #shadow and unconscious dynamics in the psyche of an AI scientist or engineer, consciously or unconsciously building bias into an AI algorithm causing serious harm to individuals and groups of people.
Geoffrey Hinton discusses the pros and cons of AI in a recent interview with PBS News.
There are very serious ethical and legal issues raised in connection with AI. As a business or organizational leader, the three major pitfalls that you must be aware of when you use AI are privacy, bias, and replication. I will focus on bias for the purposes of this blog.
Bias in Artificial Intelligence
Bias in the workplace is often talked about, and for very good reason because it can be a real problem in an AI application. For example, let's say you use your AI machine to recruit a new senior executive or manager at your company and you ask the AI machine to look for work experience patterns of current executives in order to find suitable candidates for recruitment. Now, if current executives are all white males in their 30s and 40s and you ask the machine to predict who are the most likely candidates you should approach, you're very likely to approach white males in their 30s and 40s. So you are portraying biases or building algorithms with intended or unintended biases into the future if you're not careful.
AI relies on data so if you have an unrepresentative sample of data, you will also get an unrepresentative output in your prediction or AI output. Equally, if you collect a set of data based on a specific population, if you don't update it as time moves on, your prediction may come out of date. So you can also create a bias because you're using outdated data to make your prediction. These are the two common sources of bias that are used very often about gender and in predictive policing that often impact specific racial groups.
Shadow
The unconscious can also be a real problem in developing your AI application. Jung's most direct and clear-cut definition of the shadow is 'the thing a person has no wish to be' (A Critical Dictionary of Jungian Analysis, Samuels et al, 2013). We all have a shadow, it is in fact what makes us distinctly human, we're not all good and not all bad but a combination of both.
'Everyone carries a shadow, and the less it is embodied in the individual's conscious life, the blacker and denser it is. If an inferiority is conscious, one always has a chance to correct it. Furthermore, it is constantly in contact with other interests, so that it is continually subjected to modifications. But if it is repressed and isolated from consciousness, it never gets corrected and is liable to burst forth suddenly in a moment of unawareness. At all counts, it forms an unconscious snag, thwarting our most well-meant intentions' (Collected Works 11. CG Jung. para.131)
The definition sounds like a simple statement, however, it is much more multi-layered in its manifestations as a product of the psyche. The shadow can also be described as:
the negative side of the personality
one's dark side
the other person in one
the worthless and primitive side of human nature
the unpleasant qualities one wants to hide
one's inferiority
one's undeveloped, repressed, hidden personality traits or attitudes (introversion or extraversion), functions (sensation, intuition, feeling, thinking).
Jung found a convincing explanation not only of personal antipathies but also of the cruel prejudices and persecutions of our time; biases, discrimination, hatred, and evil (A Critical Dictionary of Jungian Analysis, Samuels et al, 2013). As a very dark-skinned black man living in a predominantly Western, white society, I can tell you that I constantly attract the shadow projections of others who perceive me as aggressive, criminal, suspicious, threatening, unintelligent, stupid, and so on. I've even attracted projections from a fellow student in training as a Jungian analyst who told me that they had questions and concerns about my blog and demanded that we met to discuss their concerns. Was there projection positive or negative towards me? I don't know for sure but it isn't unusual for me to be 'questioned' about my ambitions. Having a psychological understanding of shadow projections allows you to walk away from such conflict and realize that the conflict is in the other person's mind.
#Projection is a psychoanalytic term used to describe this dynamic or complex psychological mechanism, that is to say 'attributing to others something that one cannot tolerate in oneself', for example, one's shadow, as it comes into conscious awareness is characterised by specific behaviour:
attributing to somebody else something that they are experiencing and cannot tolerate
incapable of maintaining empathy with what they experience and cannot tolerate
tendency to induce behaviourally in the other the shadow qualities that they are trying to project on the other person
effort to control the other person in order to prevent what is projected from becoming conscious, and to absolve them of the other person
(Source: Otto Kernberg, Psychoanalyst and Psychiatrist. Professor Emeritus of Psychiatry, Weill Cornell Medical College and Director, Personality Disorders Institute New York-Presbyterian Hospital, Training & Supervising Analyst, Columbia University Centre for Psychoanalytic Training and Research in Borderline, the Film)
But it is not all bad. The shadow can also be positive:
contains our potential and creativity
contains the seeds for personal growth and development
lead to the transformation of the personality
admiration of others, unconsciously projecting our own positive qualities on to others, is also a form of projection
Psychoanalysis is a process of withdrawing projections from others and re-integrating or reunifying split off aspects of our personality back into our psyche, both positive and negative. It is an arduous task but well worth the effort if you're looking to improve or transform yourself.
The concept of the shadow suggests humans are split between the light and dark sides of the human psyche. The shadow is an archetype, its contents are powerful, marked by affect, obsessional, possessive, autonomous, and capable of startling and overwhelming the well-ordered ego. It initially appears in projection when it approaches consciousness and manifests as a strong, irrational projection, positive or negative upon others.
One way the unconscious can impact the development of AI is through the intentional or unintentional input of bias into an AI algorithm. The shadow is autonomous and can act so without your conscious awareness. If one's shadow contains negative personality traits these can influence your work, as we have seen with the example of algorithmic bias above, the AI scientist or engineer has not considered the impact of the AI algorithm on others, and how it may affect particular individuals or groups. AI scientists and engineers must remember the impact on people when developing algorithm.
As they say, 'garbage in, garbage out'. If you have rubbish data, AI will simply multiply that rubbish and make a mess of any decision you're going to take on those data. Sometimes even the cleanest data may lead to a biased decision. If the AI algorithm is developing the wrong way, or interacting in a way that leads to a biased decision, well that's a problem with the algorithm and not the AI scientist or engineer. This nightmare scenario happened at Amazon when it released its same-day service for its customers. All of a sudden, the algorithm which was developed to do a good job, started mapping the white and black communities in the major US cities, the algorithm was inadvertently biased. It was rectified quickly but it is not known why the algorithm started to behave 'unethically' (Oxford AI Programme).
Confronting the Shadow
Recognizing the shadow as a living part of the psyche and that one has to learn to live with it is one of the main tenets of psychoanalysis. It is not possible to eradicate the shadow but to come to terms with it. If it is left unconscious it can operate autonomously, outside of the control of the ego.
Jungian psychoanalysis and psychotherapy aim to develop an awareness of the shadow and the situations most likely to produce shadow projections in one's personal and work life. To admit, or 'analyze' the shadow is to break its grip on the personality and bring light into what was dark, to recognize and integrate one's dark side, and to channel its energy into productive and creative endeavours that do not cause harm to others. It is an ethical task of immense courage which every individual must face at some point in their lifetime. It brings a capacity for understanding in depth oneself and others, integration of one's total identity and normal functioning of one's shadow in its creative and transformative manifestations.
In the next article in the blog series, I will delve deeper into mechanisms of the shadow using the typical personality profile of an AI scientist and engineer, using the story of Frankenstein as a metaphor for the pros and cons of AI, the impact of the singularity on humankind, and show you how intentional or unintentional bias can creep into an AI algorithm causing damage to individuals and groups.
Register on the site and be the first to read my latest blog and do get in touch if you have any questions. I would be delighted to hear from you!
Comments