Life Tips & Miscellaneous Navigation of this blog Travel and History Zen and Life Tips Economy and Business
Unconsciousness and memory
Two systems can be envisaged in our minds, one being a quick, automatic, unintentional and unconscious system, such as intuitive decision-making, and the other being a deliberate and conscious system that takes time but can be controlled, such as logical decision-making.
These are called ‘System 1’ and ‘System 2’, where System 1 is the energy-saving workings of the mind, which leads to quick decisions, but may lead to irrational thinking and behaviour, and System 2 is the one that contributes to rational thinking and handing, but requires time and effort, and sometimes misses the time to make decisions, which can This is something that can happen.
AI technology, as described in this blog, aims to take the rational decision-making of System 2 to the extreme instead of humans, and IA, as described in “Overview of Intelligence Augmentation (IA) and its application examples“, is an approach to how to connect the useful parts of System 1 and System 2. System 1 is a system of systems that is used in science and the arts, and is a system of systems that is used in the arts and sciences.
In particular, System 1 is often the source of creative activities by humans in science and art, and combining it with AI-based System 2 is expected to lead to activities that generate original ideas and performances that have never been seen before.
This section describes how System 1 works, based on the book “From the Mystery of the Mind to the Science of the Mind: The Unconscious and Memory” published by Iwanami Shoten.
In the previous article, we discussed the British experience-centred Unionism and the German Monad Theory as scientific approaches to the mind. In this article, as the next step after those trends, I will discuss the trend towards the exploration of methods to quantitatively investigate the unconscious in order to investigate the workings of the mind.
Ebbinghaus’s empirical study of unconscious memory.
The German Ebbinghaus, considered the world’s first experimental psychologist, had been conducting memory experiments on himself at his home in Berlin since 1879.
Ebbinghaus applied the psychophysical method of quantitatively examining the senses, originated by Fechner, a physicist at the University of Leipzig, to the study of memory, which he summarised in his book ‘On Memory’.
In this book, Ebbinghaus posits, as Leibniz and Herbart do in “Human Creativity and the Symbiosis with AI – On the Unconscious and Memory“, that firstly, things that have disappeared from consciousness exist as unconscious memories. Secondly, they argued that these unconscious memories influence our conscious thoughts and actions, although their existence is not directly observed, and tried to prove them by experimentation.
Specifically, in order to study pure memories that are not influenced by past experiences, they created more than 2000 pieces of meaningless learning material (meaningless spellings) such as ‘WUX’, gathered these together to create a coherent set of several learning materials, and recorded the time it took to read these out loud and memorise them completely. Next, after a certain period of time, the time taken to completely memorise the same material again was measured.
If, for example, it took 100 seconds the first time and 80 seconds or less the second time, the difference in time was called the saving rate, and data was collected showing a saving rate of 20 seconds or 20%. Ebbinghaus believed that this quantification index, the saving rate, was unconscious memory.
In this experiment, he found a forgetting station line that quantified the decrease in unconscious memory over time, and found that the savings rate (unconscious memory) decreased rapidly immediately after learning, but after a certain amount of time it became almost an asymptote and remained stable at around 30% after one month. Ebbinghaus argues that this 30% represents the amount of unconsciousness in a person’s memory.
Gorton’s demonstration of the workings of unconscious memory.
Ebbinghaus’s experiment was incomplete as an experiment in part because of the overlap in the subject data. At the same time in the UK, Goulton conducted experiments to quantify the mechanism of ideational association and wrote a book entitled ‘Studies in Human Ability and Development’.
In contrast to Ebbinghaus, who conducted experiments to minimise the influence of association, Goulton chose familiar words for which associations had already been formed, and after examining the association procedures and recording methods, he investigated the associative time until the idea associated with each word came to mind, its content, and the time and place of the association. The results of these experiments were used to quantitatively and patiently test the associations, the number of associations duplicated when the same word was used at different times and places, the distribution of the number of associations by chronological age of life, and the types of images that come to mind when the associations become angry.
These experiments have revealed that some of our past events remain in unconscious memory, and that, in addition to the various and sundry mental operations being carried out in a semi-conscious state, ideas related to consciousness arise automatically in the unconscious layer, and that these unconscious ideas are useful as material for writing and speech. The study reveals that these unconscious ideas serve as material for writing and speech.
Biological support for the workings of unconscious memory by Zeemon.
In response to Goulton’s experiments, the German biologist, Szeemon, attempted to develop a biological theorisation that extended the mechanism of memory to all organisms.
Zeemon’s breakthrough in memory was that he attempted to theorise not only conscious memory but also unconscious memory, and that he divided the memory process into learning and recall and tried to explain the whole phenomenon of memory in a unified way based on the interaction of these two processes.
In explaining these theories, Zemon used his own terms such as ‘mneme’, ‘engramme’, ‘engrafi’ and ‘ekphory’. These are ‘memory (phenomenon)’, ‘memory trace’, ‘inscription (or learning or symbolisation)’ and ‘recall (or retrieval)’ in the current field of memory theory.
According to Zemon’s mnemonic theory, an organism in the absence of any stimulus is in a kind of equilibrium state (called the primary sedentary state), and when a stimulus is added to it, the organism enters a kind of excited state and produces a specific response. If the stimulus is then removed, the organism is no longer in the same state of equilibrium (secondary sedentary state) as it was before the stimulus, so it tends to react in a similar way (excited state) under a similar stimulus. Zeemon called all such changes in the internal state of the organism engrammatic, and the whole process that causes engrammatic is called engrafi. This series of processes is the established state of learning, which Zeemon considered to occur consciously or unconsciously.
After a period of time, when a previous stimulus is given to the organism, the engraved engrammes are re-excited, causing the same response as before. This process is ectopoly and corresponds to recall or retrieval.
It is important to note that a stimulus is not necessarily a single stimulus, but often a combination of several stimuli, and ectopoly can also occur with some of the stimuli, even if they are not identical to the original stimulus. While associationism only targets recall of the exact same ideational associations as the formed ideation, in Muneme theory, recall can also take place even if only some of them match. This is similar to the story of spatial similarity as described in “On similarity in machine learning“.
Claverade’s research on the workings of unconscious memory in patients with amnesia and Schacter’s research on subconscious memory and priming.
Later, ‘behaviourism’ (an approach based on the claim that only ‘observable behaviour’ should be targeted, rather than ‘introspection-based consciousness’ as had been the case in psychology up to that point) advocated by Watson, as described in “On emotion recognition, Buddhist philosophy and AI“, dominated the field of psychology, and the term ‘memory’ was no longer even used. The term ‘memory’ was no longer even used in psychology, and much of the research on ‘memory’ came to be referred to as ‘language learning’.
In this context, research on the unconscious continued with studies of clavalate for amnesic patients with brain damage due to illness or disability, Morizon for epileptic patients, or Schacter’s study of the effects of unconscious memories (named subconscious memories by Schacter) left behind in amnesic patients on their behaviour. This will continue with research into the influence of unconscious memories (named subconscious memories by Schacter) on behaviour.
According to Schacter, everyone has unconscious subconscious memories that are completely separate from memories that are consciously remembered and recalled (healthy memories), and the effects on thoughts and behaviour that accompany the excitation (activation) of these subconscious memories are automatic and beyond the control of our conscious mind.
Such subconscious activation is termed ‘priming’, meaning ‘ignition’ or ‘ignition’, in which a preceding stimulus or experience influences subsequent thoughts and actions in the form of subconscious activation, without conscious memory.
Such priming is said to appear not only in artificial experimental settings, but also in various scenes of unconscious plagiarism, racial prejudice of which the individual is unaware, and the influence of advertising.
Priming and AI
This priming is also an interesting concept in the field of AI, and there is active research on using the concept of priming to improve human-AI interactions.
For example, in AI-enabled experience (UX) design, when a user performs a specific task, an AI assistant can more accurately understand the user’s intentions and present relevant information and context in advance, so that subsequent operations can proceed smoothly, such as Priming can be considered.
In addition, machine learning models, especially neural networks, are trained using large amounts of data, and this training data has a significant impact on the performance of the model, so the choice of data sets and their pre-processing is important. It can be said that data priming is occurring, in the sense that more exposure to a particular type of data makes it easier for the model to learn the patterns associated with it.
Understanding the principles of priming and applying them to AI systems will enable the development of more effective and user-friendly systems, and research into priming will become increasingly important as AI evolves.
コメント