Unit 1 - Scientific theories as computer simulations
Scientific theories of human behaviour and human societies tend to be expressed by using words, but words have unclear meanings: they mean different things to different people and they are value-oriented and emotionally charged.
These properties of words are a problem for science because from verbally expressed theories it is difficult to derive precise, quantitative and non-controversial empirical predictions.
And, obviously, the problem cannot be solved by defining the meaning of a word by using other words.
If verbally expressed scientific theories have these limitations, another possibility is to mimic physics and to express theories of human behaviour and human societies by using mathematics.
Unlike words, mathematical symbols have precise and unambiguous meanings and mathematics is a powerful tool not only to describe precisely and non-controversially the empirical data but also to find precise relations among the empirical data.
Are mathematical theories a possible option for the sciences that study human beings and their societies? Mathematical theories are all right when what is counted and measured is clear and all scientists agree on it. This is true for the natural sciences but it is not true when the object of study are human beings and their societies. Mathematical theories capture the regularities which exist in human phenomena but they do not explain these phenomena.
Today, computers offer a new possibility: scientific theories can be expressed as computer simulations.
A theory is translated into a computer program and if, when the program runs in the computer, its results reproduce the empirical phenomena that we want to explain, then the theory/simulation explains the empirical phenomena.
According to this new paradigm for doing science, scientists should not ask themselves “How can this phenomenon be explained?” but “How can this phenomenon be reproduced in the computer?”.
In the next slide we list five advantages of expressing scientific theories of human beings and human societies as computer simulations.
- Scientific theories as computer simulations provide operational definitions of words. Words are no more defined by using other words but they are translated into this or that feature of the computer program. If this is not possible because a word has an unclear or ambiguous meaning, the program cannot be written - and this is under the eyes of everybody.
- With respect to mathematical theories, computer simulations clearly identify what is counted and measured because what is counted and measured is specified in the computer program.
- The computer program does not only find the regularities which exist in the empirical data but it also identifies the causes of these regularities.
- The computer generates a great number of non-controversial and quantitative empirical predictions because the results of the simulation are the predictions derived from the theory/simulation.
- Computer simulations can also be experimental laboratories. After you have constructed a simulation, you can predict what will happen if you vary this or that aspect of the simulation, or the value of this or that variable and the computer will tell you if your prediction was correct.
Expressing scientific theories as simulations makes it possible to solve another problem which afflicts the science of human beings and explains why the science of human beings is so little advanced and so unsatisfactory when compared with the science of nature.
Traditionally, science is divided into disciplines but its divisions into disciplines reduces its capacity to make us understand reality because reality is not divided up into separate parts.
Reality is a very large ensemble of different phenomena but these phenomena are all connected together and, often, the phenomena studied by one discipline can only be explained by taking into consideration the phenomena studied by another discipline.
Both the science of nature and the science of human beings are divided into disciplines, but the division into disciplines is not a problem for the science of nature because physics, chemistry and biology use the same empirical methods, they have very similar conceptual and theoretical traditions and they share a view of nature as made up of causes and effects and as possessing an inherently quantitative character.
In contrast, the division of the science of human beings into disciplines (psychology, anthropology, linguistics, sociology, economics, political science) has many negative consequences because these disciplines do not share the same empirical methods, they have very different conceptual and theoretical traditions and they do not have a unified view of the phenomena they study.
The consequence is that what the science of human beings gives us are the separated pieces of a mosaic but not a unified picture of human beings.
Today, some scientists try to solve the problem by doing inter-disciplinary research. Scientists of different disciplines collaborate together to find the relations that exist among the phenomena studied by their different disciplines. But inter-disciplinary research does not really solve the problem because each discipline has its own theories, and science is not only made of empirical phenomena but also of theories that explain the phenomena.
The problem can only be solved by expressing scientific theories as computer simulations. Computer simulations are a unified theoretical and methodological framework, a “lingua franca” that facilitates the dialogue among the different scientific disciplines and the progressive development of an integrated, non-disciplinary science of human beings. The brain of a scientist is too small to be able to formulate a theory which explains the phenomena studied by so many different disciplines.
But if the theory is expressed as a computer simulation, the theory is in the computer, not in the scientist’s brain, and the computer can store enormous quantities of data and it can compute what happens when all the data interact together.
The next three slides schematically represent disciplinary research, inter-disciplinary research and computer-based non-disciplinary research.
(a) The small circles indicate what each of five scientists (black squares) directly knows about reality if he or she works alone.
(b) The large circle indicates what the same five scientists know about reality because they belong to the same discipline and they collaborate together. The area of the large circle is much greater than the sum of the areas of the five small circles and this explains the importance of scientific collaboration and of scientific disciplines.
(a) Quantity of knowledge possessed by fifteen scientists who belong to three different scientific disciplines.
(b) Quantity of knowledge possessed by the same number of scientists if the three different disciplines collaborate together and do inter-disciplinary research. Inter-disciplinary research increases what science knows about reality.
Expressing theories of human beings as simulations has another important advantage, this one of a different nature. All human beings have values, desires and fears, and these values -desires and fears- make it difficult for them to understand themselves as science understands all other phenomena of reality because they tend to confuse what they would like to be with what they actually are. Since scientists also are human beings, they almost inevitably carry their values, desires and fears with them when they do science. But science can really know and understand reality only if scientists look at reality with complete detachment and with a mind free from pre-conceptions and values.
This is easy to do when scientists study nature but it is almost impossible when they study themselves - and this is another explanation why the science of human beings is so less advanced compared to the science of nature. Expressing theories of human beings and human societies as computer simulations solves this problem. If a theory is formulated verbally, scientists tend to derive from the theory only those predictions that correspond to their desires and values. If a theory is expressed as a computer simulation, scientists are necessarily confronted with all the predictions that derive from the theory, both those that correspond to their desires and values and those which are in contrast with their desires and value.
Unit 2 - Robots as scientific theories of human beings
Understand human beings
So, if we want to understand human beings and their societies as science understands all other phenomena of reality, we must simulate human beings and their societies in the computer.
Today, there are various attempts at constructing computational artefacts which have a “mind” and are “intelligent” but these artefacts do not have a “body” and, since behaviour is the result of the physical interactions of the animal’s body with the physical environment, to make us really understand human behaviour, these artefacts must not only have a “mind” and an “intelligence” but they must also have a “body”.
A simulated or physically realized computational artefact which has a body and behaves like a human being - or some other animal - is a robot.
Understand human beings
Robotics is a very active area of research today. Many people are busy constructing all sorts of robots and robots are a popular theme in the media.
But what really are robots? As we have said, robots are computer-based artefacts which physically resemble an animal or a human being and behave like an animal or a human being.
This is an acceptable but, for our purposes, insufficient definition because, if we want to really know what are robots, it is necessary to explicitly declare why we construct robots.
Why construct robots?
Robots can be constructed with two different purposes.
They can be constructed because they have practical applications and economic value - robots as technology - or they can be constructed because they can help us to better understand human beings and other animals - robots as science. It is important to distinguish between these two different goals for constructing robots because the two goals push robotic research in different directions and favour different methods.
The applied goal and the scientific goal have, and should have, reciprocal links because robots that have practical applications can be a test for scientific theories and can suggest new hypotheses and pose new problems for robots as science, and robots as science can make it possible to construct better practical robots and can suggest new applications. But the two goals are different and they should be kept separate because, in the two cases, success is measured by different criteria.
For robots as technology (humanoid robots), we should ask: does this robot have useful applications and economic value? For robots as science (human robots), we should ask: does this robot make us understand better human beings?
Why construct robots?
This is not what is happening today. When one constructs a robot, it is not always clear if the goal is to produce some practically useful artefact or to better understand human behaviour and the behaviour of other animals.
This is a problem especially for robots as science because today, almost all research money is for robots as technology and, therefore, even researchers who are not particularly interested in the practical applications of robots tend to be guided in their work, perhaps unknowingly, by practical applications.
This is bad because robots have a great potential for our understanding of the behaviour of human beings and of the organization of their societies and the emphasis on robots as technologies is an obstacle to exploiting this potential.
Why construct robots?
If all research money is for robots as applications, the selection of research topics is guided by the possibility of applications, and this inevitably leads to ignoring many important behavioural and social phenomena that we might better understand by constructing robots that reproduce those phenomena.
Who will construct robots that make mistakes, robots that have physical or psychological pathologies, robots that sleep and dream, robots that do what they do to satisfy their, and not our, motivations, robots that really feel the emotions that their express, robots that can be intelligent but also stupid, robots that have different personalities, robots that have an unpredictable behaviour, robots that can attack other robots, robots that can organize themselves in societies? These robots would not have practical applications (at least for now) and they might even pose social problems and elicit undesirable emotional reactions. But since human beings are and do all these things, if we want to understand human beings by constructing human robots, we should reproduce all these things with robots.
Why construct robots?
Another reason why current robots tend to robots as technology rather than robots as science is that robots are physical artefacts controlled by a computer or simulated in a computer.
Therefore, engineers and computer scientists are an important and necessary component of the community of roboticists. But engineers and computer scientists are trained to construct artefacts that solve practical problems and to design systems that behave “optimally” with respect to some desired objective.
Robots as science do not have practical objectives and they do not behave optimally. They are “pure” research tools and only some engineers and computer scientists want to become “pure” scientists.
Why construct robots?
The prevailing interest in robots as technology explains why so much research in robotics is dedicated to the physical body of the robots, to the morphology and dynamics of their body, to the materials that make up their body, and to the sensors and effectors of their body.
This research is important even for robots as science because, as we have said, robots presuppose an “embodied” conception of the “mind” according to which the “mind” is primarily the result of the interactions of the human body with the physical environment. But the robots’ body can be very simplified with respect to the body of human beings and it may not be the principal emphasis of research.
And - what is more important - while robots as technology must necessarily be physically realized, this is not true for robots as science. Robots as science can only be simulated in the computer and still they may be very useful tools for understanding and explaining the behaviour of human beings and the organization of their societies.
Why construct robots?
Another important difference between robots as technology and robots as science has to do with how robots are constructed. Most of today’s robots are programmed by us to do what they do, and this may be appropriate for robots as technology. If we are interested in some particular application, we program the robot so that the robot does what it is expected to do. But programming the robots does not make much sense for robots as science. Robots as science should resemble real animals and real animals are not programmed by anyone; they autonomously acquire whatever behaviour they possess.
Therefore, robots as science should not be programmed by us, but they should acquire their behaviours as a result of biological evolution in a succession of generations or learning in the course of the life of an individual robot - or, even better, both evolution and learning.
Why construct robots?
But even robots that acquire their behaviour through evolution or learning can be robots as technology rather than robots as science. If our goal is to construct robots that possess some specific behaviour, we evolve the robots or make the robots learn so that at the end they exhibit the desired behaviour.
On the contrary, if we are interested in robots as science, we want to know which is the behaviour that they acquire in a given environment, how they acquire the behaviour, what are the different components of the behaviour, what other behaviours they also acquire, how different is the behaviour of the different robots, why some robots do not acquire the behaviour.
This course is dedicated to robots as science, not to robots as technology. As I have already said, there should be a dialogue between robots as science and robots as technology because the dialogue may be reciprocally useful. But, unless we say explicitly that we are interested in robots as science, not in robots as technology, one may not understand why most of the robots described in this course have been constructed.
Unit 3 - Human and humanoid robots
Robots as scientific theories are something new and we are still trying to understand how to use robots to explain human beings and human societies.
Robots simplify with respect to real human beings and real human societies but this is not a problem because scientific theories must simplify with respect to reality.
Scientific theories are not a re-description of reality. They must capture the basic entities, mechanisms and processes that are behind what is observed and explain what is observed.
However, robots may be only “toys” which do not tell us much about human beings and, to avoid this danger, we must adopt a general principle of science: “one theory/many phenomena” - which in our case becomes “one robot/many phenomena”. A robot should not reproduce one single behavioural or social phenomenon but the same robot should reproduce as many different behavioural or social phenomena as possible. If a robot reproduces only one phenomenon, the robot is likely to be only a “toy”.
If the robot reproduces many different behavioural and social phenomena, we may be more confident that the robot tells us what human beings are and how human societies function. Why?If a robot reproduces one single phenomenon, it is possible to construct many different robots that reproduce the same phenomenon, and deciding among the different robots will be arbitrary.
If one and the same robot reproduces many different phenomena, it is more difficult to construct different robots which reproduce all these phenomena together and, therefore, the robot will more likely tell us something interesting about human beings and human societies. The slides of this Unit list a certain number of different phenomena that the same robot should be able to reproduce.
Psychologists tend to study the “mind” without taking the body into consideration and only recently - partly, under the influence of robotics - embodied theories of the “mind” have made their appearance. The body is important to explain behaviour because behaviour consists in the movements of the different parts of the body in response to the stimuli that arrive from the body’s sensors and, therefore, animals which have different bodies cannot have the same “mind”.
Unlike “disembodied” computer-based artefacts such as those of artificial intelligence, robots are a first step in the direction of “one theory/many phenomena” because they reproduce not only behaviour but also the body.
In fact, robots are the fourth revolution in how we view human beings.
The first revolution was the Copernican revolution: the place where human beings live, the Earth, is not the centre of the Universe.
The second revolution was the Darwinian revolution: human beings have not been created by God but they have been created by nature.
The third was the Freudian revolution: human behaviour is not guided by reason but by motivations of which we may not even be conscious.
Now a fourth revolution has begun, the robotic revolution: human beings are bodies.
Internal organs and systems
But the body should not only be the external body - its shape, its size, and its sensory and motor organs. The body should also be the organs and systems which are inside the body.
One particular internal organ, the brain, is especially important because it directly controls behaviour.
However, behaviour is not only under the control of the brain but it is also influenced by other internal organs and systems such as the heart, the lungs, the gut system, the hormonal system, and the immune system.
Therefore, the body of our robots should contain all these organs and systems.
Behaviour does not only depend on the organism but it is a result of the interactions of the organism with the external environment.
Psychologists have difficulty recognizing that to explain the behaviour of an organism it is necessary to take into consideration the environment in which the organism lives.
This is due to the historical traditions of the discipline but also to practical or technical reasons because studying the behaviour of an organism in its environment is difficult and expensive in terms of both time and money, and in many cases impossible.
But if we want to apply the principle “one robot/many phenomena”, our robots should not only be artificial bodies but they should also live in a simulated “natural” environment and they should be free to do whatever they want to do in this environment. Robotics as science must be an ecological robotics and we should also do “ecological experiments” in which we vary the environment in which the robots live and see the consequences of these environmental variations for their behaviour.
We should also do “laboratory experiments” with our robots.
We extract a robot from its natural environment, we bring the robot into a simplified environment in which everything is controlled by us and we collect precise and quantitative data on the robot’s behaviour in this controlled environment.
Evolution and learning
This course is based on the idea that, if we want to understand human beings, we must construct robots that reproduce human beings.
But there is a problem with the idea of “constructing” robots. Human beings are not constructed by anybody. They are the result of various historical processes (evolution, development, learning, culture and history) which have made them what they are. Therefore, if robots must replicate human beings, they cannot be constructed by us but they must construct themselves in the course of these different historical processes.
As we have said, robots are an example of a new, general approach to doing science which is based on the principle “If you want to understand X, you must simulate X in a computer”. This principle now becomes “If you want to understand X, you must simulate how X has become what it is”.
Human beings are different from non-human animals because each species of animals is different from all other species.
However, to better understand human beings it may be useful to compare them with other animals.
The comparative approach should be extended to robotics. We should construct not only human robots but also worm robots, fish robots, mouse robots, and monkey robots, and we should do a “comparative robotics” because a comparative robotics can be as useful to robotics as comparative biology and comparative psychology are useful to traditional science.
But there is something that we can do with robots and only with robots. We can construct counterfactual robots, robots which are like possible animals but not like any existing animal. These robots are counterfactual hypotheses, hypotheses about possible but non-existing worlds.
The extension of the comparative method to non-existing worlds which is made possible by expressing theories as computational artefacts is an important addition to the tools of science. Before the computer, counterfactual hypotheses could only be ideas expressed and discussed in words.
The computer changed this. Counterfactual hypotheses have become worlds which do exist in the computer or in a computer-controlled physical artefact and we can observe these worlds and do experiment with them as we observe and do experiments with the real world.
Risorse della lezione
- Quiz: Quiz Lesson 1 - Science
- Quiz: Quiz Lesson 2 - Motivations
- Evolution and learning
- Quiz: Quiz Lesson 3 - Evolution and learning
- Quiz: Quiz Lesson 4 - Language
- Mental life
- Quiz: Quiz Lesson 5 - Mental life
- Quiz: Quiz Lesson 6 - Families
- Quiz: Quiz Lesson 7 - Cultures
- Quiz: Quiz Lesson 8 - Economies
- Political life
- Quiz: Quiz Lesson 9 - Political life
- Inter-individual differences, mental pathologies, art and religion
- Quiz: Quiz Lesson 10 - Intern-individual differences, mental pathologies, art and religion
Immagine slide 2
Immagine slide 3
- Scientific theories are expressed by using either words (verbal theories) or mathematical symbols (mathematical theories).