Your complimentary articles
You’ve read one of your four complimentary articles for this month.
You can read four articles free per month. To have complete access to the thousands of philosophy articles on this site, please
Antoni Diller isn’t one. And he can prove it.
We are familiar with the depiction of androids in popular science-fiction films and television programmes. Arnold Schwarzenegger, for example, in the film Terminator (1984), portrays an android who is sent back in time in order to assassinate the mother of the human leader in the war against the machines. Lieutenant Commander Data, an android, is a regular member of the crew of the Starship Enterprise in the cult television series Star Trek: The Next Generation. Usually, androids in popular science fiction are easily distinguishable from their human colleagues. Sometimes the tell-tale signs, such as Data’s inability to use contractions, are laughable, but other differences are more significant. Data, for example, cannot experience or express emotions. It would be easy, however, to write a science-fiction story in which androids indistinguishable from human beings featured, but it would also be largely pointless.
Although humanoid robots do not currently exist, scientists at several research laboratories around the world are trying to make such robots. (One such venture, called the Cog Project, is taking place at MIT’s Artificial Intelligence Laboratory. More information about this can be found on the laboratory’s web site at http://www.ai.mit.edu/projects/cog.) If any of these groups succeeded in manufacturing an undetectable android, that would tell us something important about human beings. It would prove that we are nothing but machines whose behaviour can be completely explained in scientific terms. The question whether an undetectable android could be constructed some day, therefore, is not merely a technical or scientific problem. It is also an important philosophical one. In what follows I am going to assume that the engineering aspects of the project can be overcome, and I will concentrate instead on some of its philosophical aspects. I am assuming, for example, that sufficiently powerful miniature motors can be made in order to move the android’s limbs and that electronic sensors can be devised which mimic the way in which human senses work. The conclusion that I am arguing for is that it will always be possible to detect an android that has been designed and built by human beings. There will always be some behaviour that distinguishes a manufactured android from a real human being.
Before looking at the details of the argument, I will present its outline. In order for an android to be built it first has to be designed and the designer or design-team has to make use of a large number of scientific theories. The android’s behaviour is produced by means of these theories and so she will behave in accordance with them. The psychological theories used not only describe the behaviour of the android, but they also produce it. In the case of human beings, however, psychological theories are only attempts to describe and predict their behaviour. It is impossible for us to know with certainty how that behaviour is produced. It is this difference between humans and androids that means that there will always be behaviours that allow us to differentiate between them.
Having presented the argument in outline I will now fill in some of the details. A large number of theories have to be incorporated or embedded in the android in order to try and get her to produce the full repertoire of human behaviour. To be potentially undetectable the android would have to be able, for example, to express emotions appropriate to the various circumstances in which she found herself. She would also have to be able to communicate verbally with human beings about a wide range of topics and act appropriately in different settings such as restaurants, banks, pubs, post offices, libraries, churches, cinemas, casinos and so on. She would also have to be able to do a large number of other things. To make the argument easier to follow I will concentrate on just one type of behaviour. I will restrict my discussion to behaviour which is associated with emotional states in human beings. Furthermore, I will assume that a single theory is responsible for producing emotion-related behaviour in the android. In reality, several complementary and, hopefully, mutually consistent theories would probably be used, but for the purposes of my argument these can be thought of as having been combined into a single theory. It should be noted that I am not here concerned with issues relating to whether or not the android could experience anxiety, say, in exactly the same way that a person does. Nothing that I say depends upon the felt quality of different emotions. I am solely concerned with the external manifestations of some emotions. In the case of extreme anxiety, for example, these might include trembling, shortness of breath, excessive sweating and difficulty in swallowing. In addition, I am neither assuming that every emotional state experienced by an adult has some distinctive behavioural counterpart nor am I assuming that all expressions of emotion are sincere. To be Detecting Androids Antoni Diller isn’t one. And he can prove it. Winter undetectable the android would have to be able to hide her emotions on some occasions and also to falsely express emotions that she was not experiencing on other occasions. It should be noted that with suitable changes the argument that I am presenting would apply to all the different kinds of behaviour in the android’s repertoire and not only to emotionrelated behaviour.
Our current scientific theories, although very good approximations to the truth, are always open to improvement. Strictly speaking, this means that they are false. Furthermore, as they have replaced the theories of the past, we are led to the position that all scientific theories are false. Not all false theories are useless, however. Einstein showed, for example, that Newton’s mechanics is false and yet it is such a good approximation to the truth that it was used to calculate the paths that rockets sent to the Moon should take and, as we all know, they got there. Because the claim that all scientific theories are false is counter-intuitive I will, however, say more about it later, but for the time being it will be assumed to be true.
The way in which scientific theories are falsified is well understood. A theory about the motion of the planets, say, describes their behaviour and it makes predictions about their positions in the sky. If the actual behaviour of a planet deviates from what the theory says or if a prediction turns out to be false, then the astronomical theory is said to have been falsified. As I am assuming that all scientific theories are false, the one that was used in the design of the android in order to produce her emotion-related behaviour must be false. This means that there is some human behaviour that falsifies that theory, but as the android’s behaviour is produced by the theory it can never falsify it. Any human behaviour which falsifies the theory can be used to distinguish the android from a human being, because in the same circumstances the android could not exhibit that behaviour.
Consider, for example, the case of anxiety. (To be manageable this example has been made very simple. Although not entirely realistic, it illustrates the point that I am making.) The android could have been designed using a cognitive theory of emotion. Such a theory might say, for example, that people feel anxious and express anxiety-related behaviour if they have an unrealistic demand about some impending, personally meaningful event and the more strongly that the demand is held the more severe is the anxiety that is experienced. Faced with an imminent job interview and believing very firmly that she must get this job the android would feel anxious, tremble slightly and sweat more than normal. A human being facing a job interview and holding that same belief equally firmly might, however, experience some mild apprehension and neither tremble nor sweat more than usual. If this were to happen, then we would begin to think that the cognitive theory of emotion, although on the right track, is false. We would try to improve the theory and replace it with a better one which preserved the successes of the earlier theory. Because of the way in which the android was designed, however, every time she faced a job interview believing that she had to get the job in question she would exhibit the anxiety-related behaviour already mentioned.
The behaviours which allow us to detect the android are those which falsify the theories that were embedded in her. In general, different behaviours would falsify different theories. Such falsifying behaviour cannot be exhibited by the android because she can only behave in accordance with the theories incorporated in her that produce her behaviour. I will present an analogy in order to clarify the above argument. An orrery is a clockwork model of the solar system. An orrery could be built to illustrate Ptolemy’s celestial mechanics with the Earth at the centre and the Sun and other planets moving around it in orbits that are produced by combining several circular motions. Alternatively it could be built to illustrate Copernicus’ theory in which the orbits of the planets are again obtained by combining several circular motions but now the Sun is at the centre of the system. Yet another orrery could be built to illustrate Kepler’s celestial mechanics. In this model the planets would move in ellipses around the Sun which would lie at one of the foci of each ellipse. In practice, an orrery cannot be built which is an exact scale model of the solar system. (One reason why this is the case is because the highest common factor of the mean distances of the planets from one another is very small in comparison with the mean distance of the furthest planet from the Sun.) A computer model could, however, be constructed. Imagine such a model which was built to illustrate Ptolemy’s celestial mechanics. The behaviour of the model could not deviate from that described by Ptolemy’s theory. No matter how many observations we took of the miniature solar system they would always be in conformity with that theory. It would be impossible for any observations of that model to falsify Ptolemy’s theory. We know, however, that the behaviour of the actual solar system is different from what we would expect on the basis of Ptolemy’s theory. The behaviour of the solar system that falsified Ptolemy’s astronomy could not be produced by the computer model.
In this analogy the computer model or orrery corresponds to an android and the behaviour of the model corresponds to the android’s emotion-related behaviour, the real solar system corresponds to a human being and its behaviour corresponds to human emotion-related behaviour. Furthermore, Ptolemy’s celestial mechanics corresponds to the theory of emotion that was used in order to produce the android’s emotion-related behaviour. There are discrepancies between the behaviour of the android and human behaviour in the same circumstances just as there are discrepancies between the behaviour of the model and the behaviour of the solar system. In the case of emotion-related behaviour these discrepancies allow us to tell androids apart from humans.
The premise in the argument presented above that is likely to be seen as most controversial is the one that states that all scientific theories are false. A vast amount could be said about this premise and lack of space forbids a thorough discussion, but I will try to show both inductivists and noninductivists that it has, at least, some plausibility.
An inductivist is someone who believes in some form of induction where this is a form of inference from singular or particular premises to a general conclusion. Thus, after observing a few swans and noting that they are all white, the inductivist concludes that all swans are white. An inductive inference, however, does not guarantee the truth of its conclusion (even if all of its premises are true). Although it was accepted for a long time that all swans are white, the discovery of black swans falsified this conjecture. Anyone who has any knowledge of the history of science knows that it is littered with false theories. To mention just a few: the theory of spontaneous generation; the caloric theory of heat; Aristotelian mechanics; the theory that light moves infinitely fast; Ptolemaic astronomy; the theory that the chemical atom is indivisible and the phlogiston theory. Furthermore, we cannot be certain that any of our current scientific theories are definitely true. Therefore, by induction we can argue that every scientific theory is false. This conclusion would be falsified if a theory known to be certainly true were to be exhibited. If that were to happen, it would be a very remarkable event. It is doubtful, however, that we can be certain about any general theory. This has been called the pessimistic induction, but optimistic induction would be a better name. Although all scientific theories are false, if we look at a series of theories in a given area we almost always find that later theories are much better than earlier ones. Thus, in astronomy, Einstein’s theory was an improvement on Newton’s which was better than Kepler’s and that in turn was more accurate than Copernicus’s which itself was a great improvement on Ptolemy’s. The belief that all theories are false is optimistic because it encourages the search for everbetter theories.
A non-inductivist is someone who does not accept the general validity or usefulness of induction. The best-known non-inductivist is Popper. He argues that science is a progression of bold, speculative guesses. The origins of the conjectural theories of science are irrelevant. What is important is that they are subjected to searching examination and attempts are made to falsify them. The ones that survive this rigorous testing process are tentatively accepted by the scientific community. This view of the growth of scientific knowledge presupposes the assumption that all scientific theories are false. A scientist would have little incentive to try and falsify the theories of his rivals if he thought that they were certainly true.
Obviously, a great deal more could be said about the hypothesis that all scientific theories are false, but I hope that I have, at least, shown that it is not entirely unreasonable.
In this paper I have shown that it is impossible for human beings to design and build an undetectable android. The heart of the argument depends on some properties of human knowledge. In the past some people have argued that computers would not be able to appreciate beauty, for example, or that they could not fall in love. In principle, I can see no reason why this should be the case, but what I have shown is that android aesthetic appreciation, for example, will always be distinguishable from human aesthetic appreciation and that android love could never be identical to human love. If the writers of popular science fiction are correct in some of their speculations, then androids will one day live with us and work alongside us but there will always be an unbridgeable gap between us.
© Dr Antoni Diller 1999
Antoni Diller teaches at Birmingham University.