Preface

Abstract

Schrödinger’s cat as an active player: the role of an observer. Is the human brain the mirror or the creator of the world?

Physics experienced a situation where the object of research is not independent of the observer long ago, when quantum theory began to develop and the world was confused by Schrödinger’s cat. Stories that violate all the usual ideas about space and time, as is commonly thought, are inhabiting the quantum world, while in the macrocosm, it would seem, such phenomena were not observed... But in the sciences that study living systems, the role of the observer should not be underestimated.

In the twentieth century, with the discoveries in quantum physics, it became clear that the so- called third-person science was losing its usual absolute value. Following a classic scheme, a scientist is a spectator sitting in a theatre chair, while the world they are studying is on the stage, and the spectator (that is, the scientist) does not interfere with it, “preserving objectivity”. But over the past hundred years, it has become clear that the researcher is in the same world, there is no division into stage and hall, and the fact of observation is an active component of the en- tire process. This was most vividly described in a thought experiment proposed by one of the founders of quantum physics, Erwin Schrödinger: a cat is placed in a closed box containing a ra- dioactive core and a container with a poisonous gas; if the core disintegrates (50% probability), the container opens and the cat dies. According to the laws of quantum mechanics, if no one is watching the core, then its state is described by mixing two states — a disintegrated core and a non-disintegrated core, which means that the cat sitting in the box is both alive and dead at the same time. If you open the box, you can only see one state: the core has disintegrated — the cat is dead, or the core has not disintegrated — the cat is alive.

Today we understand that the role of the observer, non-deductible from the information pro- cessing, is relevant not only for the quantum world. This was written by major scientists and philosophers — from L. Wittgenstein, who introduced the metaphor of the text as a carpet from which everyone pulls their threads, thereby “reading” it in their own way, to P. Floren- sky, N. Berdyaev, A. Ukhtomsky, A. Pyatigorsky, M. Mamardashvili, V. Zinchenko, A. Leontiev,

V. Lectorsky...

We only see what we know. Images and representations are not a copy or sum of the physical sig- nals sent to our receptors. Our brain builds them. In other words, what is seen, heard and felt is different not only in different animal species and in all of them from us, not because the ranges of vision, hearing, smell, and so on are different, but because all creatures have their own brain, which processes sensory signals, forming subjective (!) images. They are different not only in different species but also in different people. Therefore, it is time to learn to make serious ad- justments for individual, ethnic, religious, professional and other cultural differences that built the neural network and subjective worlds of different people. N. Marr, I. Frank-Kamenetsky, O.

Freudenberg, and P. Gurevich described the cardinal importance of not only individual but also cross-cultural specifics of the thought process...

The brain is not the sum of billions of neurons and their connections but the individual experi- ence that shaped this tool and tuned it. Perception is the active extraction of knowledge and the construction of the world. Different living systems do this in their own way, “pulling the right threads out of the carpet”, for example, magnetic fields or polarized light, and building specific models of the world.

All cognitive processes consist of receiving and processing information according to certain rules and algorithms. The brain forms mental representations – hypotheses of a high degree of abstraction that underlie the picture of the world, which cannot be verified empirically because there is simply no “objective” or “real” picture of the world [Chernigovskaya, 2010-2017; Alek- sandrov, 2009-2018].

It is clear that the processes of working with memory (writing, reading, searching) in humans and computers are very different. The organization of computer memory is based on addressing, that is, indicating the location of information in the memory; various types of content search (by keys, feature sets, etc.) are provided by a system of address links. Human memory also has a large set of keys that allow us to quickly read the necessary information. However, even if we get comparable results, we have no confidence that the processes themselves were the same! For ex- ample, there are robots that can compensate for movement disorders by continuously re-mod- eling themselves depending on the situation. Does this mean that the robot has self-awareness and subjective reality, free will to make decisions about itself?

Context as an active player, or one is not equal to one: ambiguity resolution

People live in a world where they are constantly faced with vague and ambiguous information. However, they must make decisions by decoding information as unambiguously as possible and relevantly to the situation. Ambiguity and blurring of information concern all modes of percep- tion, but this is especially evident in the example of human verbal language.

It would seem that if the main function of a language was communication, then the ambiguity in such code should have been eliminated as quickly as possible. However, in reality, we see a completely opposite picture. Perhaps we should once again listen to N. Chomsky, who believes, among other things, that language is not so well adapted for communication, and was formed mainly for structuring thinking, that is, for “internal” processes; the communicative function, in this case, is sort of a by-product. Verbal language provides the nomination of mental representa- tions coming from sensory input and thus “objectifies” individual experience, providing a de- scription of the world and communication. However, studies on communication theory have long discussed the so-called communication pits, gaps in understanding, very often formed despite the seemingly correct construction of the message (see, for example, numerous works:

H. N. Clark, H. Sacks). Thus, ambiguity and multiple meanings should be reduced to zero in any type of communication — so that ideally each word or construction has a single meaning. Similarly, it would be reasonable to expect that languages would get rid of ambiguity as they

evolved, but this also contradicts the facts. For example, T. Gibson believes that ambiguity in verbal language represents an advantage and a means of saving, since the same words can be used in different situations and with different meanings, and a specific situation is resolved using context: corpus studies in several languages show that shorter and more frequent words are the most polysemous, which confirms the idea of the parsimonious lexicon. There is also a psychological explanation: it is believed that the speaker cognitively “profits” from conveying more information with less, while the listener “profits” too from including all types of contexts to decode a compact message correctly, instead of having to analyze compositionally and syn- tactically complex constructions.

It is particularly interesting to investigate the mechanisms of ambiguity and multiple meanings in works of art, where there is an opposite task of increasing the number of options for under- standing and reading, rather than reducing them. This area is clearly insufficiently developed within the framework of cognitive science (see [Bacha-Trams et al., 2018; Apanovich et al., 2018; and others])

It is hard to argue with the fact that the intuitive, metaphorical cognitive tools with a different logic are no less powerful than classical logic and its consequences. As beautifully defined by D.

S. Chernavsky, the transition from intuitive to logical is a process of transfusion of information from one container to another, less capacious and more rigid. Some information is lost during that process. The value of the lost information depends on the purposes for which it could be used. One cannot but agree that a logical description of the world can become an obstacle to obtaining new knowledge that does not correspond to the usual rules. Of course, logic has evolved, approaching more and more in its various guises what we are accustomed to consider- ing as real world, and the most effective in this way, of course, is fuzzy logic.

The significance of ambiguity research has applications for a number of sciences –psychophys- iology and sensory physiology (perception of aural, visual, and especially tactile and olfactory information), for cognitive psychology, in particular the psychology of perception of visual and auditory representations, for information theory and research on natural language processing by humans and in automatic analysis, for the investigation of the effectiveness of human-com- puter systems. Modeling ambiguity resolution algorithms in artificial neural networks is of par- ticular interest. There are also many practical applications – from psychological and linguistic examination to medicine and artificial intelligence.

The perceiver of speech (listener or reader) constantly makes predictions based on the existing context, anticipates what has not yet been heard or read. Our task was to describe this process: on the basis of what these hypotheses are put forward, how they are verified, what happens to correcting if the perceived speech material (word, sentence, text) turns out to be ambiguous.

There are several approaches to describing ambiguity resolution strategies. The sequential pro- cessing model assumes that interpretations do not occur simultaneously, but in turn, with the first (“preferred”) interpretation chosen based on certain principles, and only subsequent ones, if the first one fails, are made based on the context. The parallel processing model is based on the

concept of competition: it is assumed that initially when a text is perceived, all possible variants of its interpretation are considered at once; they compete with each other, and eventually, one of the variants is assigned more weight than the others and is recognized as correct. The delayed processing model assumes that decision-making is deferred until a resolving context appears, meaning that an ambiguous fragment is not interpreted until the necessary information is ob- tained. This book provides a systematic description of what linguistic and psychological factors may determine the choice of a particular strategy for speech processing at different linguistic levels: when working with words, sentences, and texts.

The extraordinary role of context, and hence the possibility of multiple interpretations of mes- sages and events in general, is a cardinal characteristic of information processing at all levels.

All of the above points to the fact that in order to perceive, process, and transmit information, we need to know the characteristics of the subject. Our virtual assistant — the Schrödinger’s cat — is a model of such a subject, and we study how it perceives the world...