The Starting Point

Although we come into the world with no knowledge of our surroundings or of who or what we are, we have a collection of instinctive behaviours and responses to stimuli applied to our senses, or reflexes to get us started.

These include the sucking reflex and the instinct to put out one’s hands to break a fall and reduce the likelihood of injury.

Although we have no direct experience before we are born of the environment in which we will have to function, our brains are to some extent structured so that we are able to learn how to do so. A kind of BIOS and some specialised hardware but very little else.

In his book “Descartes’ baby”, Paul Bloom [1] quotes some properties of the external world which appear to be built in to our brains at birth which are listed by Spelke 1994 [2].

• Cohesion: If a hand pulls at an object, babies expect the entire object to go off with the hand. If it comes off in pieces then they are surprised showing an expectation that objects are cohesive.
• Continuity: Imagine a stage with two vertical barriers separated in space. A small object, like a box, goes behind the barrier on the left, continues between the barriers, goes behind the barrier on the right and comes out on the other side. Adults see this as a single object and so do babies. Now suppose that the box goes behind the barrier on the left, there is a pause, and then the box emerges from the screen on the right, never appearing in the gap. Adults assume there are two boxes here, not one. Babies make the same assumption; they expect continuity.
• Solidity: If an object is put immediately behind a screen, and then the screen tilts backward, babies expect it to stop moving – it should hit the object. When it goes through the space that should be occupied by the hidden object (a trapdoor is used) babies look longer. They expect objects to be solid.
• Contact: One object heads toward another, but the second object moves away an instant before the first object hits it. For babies, just as with adults, this action-at-a-distance is surprising; it violates our expectation of contact – that objects can only influence each other by touching.

In addition to that, Bloom [1] identifies several innate traits which also facilitate learning and modelling the environment.

• Essentialism: The tendency to categorise objects, animate or otherwise, into groups which have common properties. To recognise that objects may be in the same group due to what is inside rather than just outward appearance. [page 47]
• Promiscuous Teleology: A tendency to attribute purpose to every object including those which have formed by accident. [page 62] They will generally ascribe design to anything both artificial and natural.
• Creationism: The tendency to expect that if something exists then it must have been intentionally made. “to a 4 year old everything looks as if it has been created for something.”[page 62]
• Dualism: Babies treat people differently from objects. In fact they can be “trigger happy” in attributing mind to patterns which are associated with people.

From about age 2, the baby starts to learn, by imitation and by habit forming to see what works and what does not. This is generally automatic and without intention. Later on the child will actively investigate the environment and come to conclusions about it, firstly by means of induction and, later on, also by deduction. The child, and later the adult, will ask questions such as “Why does this happen?”, “How does this work?”, “How did this come about?”, “What is the purpose, or function, of this?” and will try to find out by experiment or by consulting authorities such as parents, teachers or books.

Knowledge from empiricism.

From this starting point, the baby shows an “exploratory urge” to make sense of the world and to predict how its environment behaves under various conditions and how it responds when prodded. Mental models are built up which get more accurate and more expansive as time goes on. For instance the expectation that an unsupported object will fall under gravity in different contexts is known by the age of 3-5 months [page 13]. The attribution of agency, or “mind”, to other people and objects happens by about 1 year old [page 17], false belief tests, the ability to realise that people will act according to what they believe to be true rather than what actually is true, about 4 years [page 20], learning cultural taboos [page 23] and so on.

So, as individuals grow up, in the absence of brain pathology, more and more models of the world are learned, particularly those which enable us to effectively communicate with others and to learn from them. In the early days, this is done unintentionally and unconsciously and relies on the in-built properties of the infant brain. Later on, this process is done intentionally and consciously driven by curiosity or necessity and either by conducting experiments or by asking trusted authorities.

This active experimentation to probe the environment and see how it responds to different actions can confirm or correct the models we have formed and enable us to form new ones. So long as our environment responds in a consistent way, which is usually found to be true, we can formulate rules and strategies about “what works”, or what is likely to be necessary in order to produce a desired outcome. This process is a precursor of the “scientific method” which can lead to description, prediction, explanation and manipulation of the observed environment.

Agents and non-agents

Our observations of the world tell us that, most of the time, things are predictable albeit not always in a simple way. The same experiments with the same initial conditions will always yield the same outcome. However, we also observe that the world contains highly complex structures whose behaviour cannot be predicted with certainty. Instinctively, we categorise these entities into “agents” and “non-agents”. Generally, those entities who are perceived to look like us and behave in a similar way to us will be assumed to think like us. They will be considered as agents.

The behaviour of agents, like ourselves, is, at least partially, the result of free choice and for that reason not predictable, even in principle, by anyone else. A system may be complex enough so that it’s behaviour is unpredictable to us, such as the weather system or even a chaotic pendulum, but that it is deterministic, so that given enough computational power the behaviour can be predicted or random, either in the sense of the initial conditions being unknowable or truly random as in quantum effects. Nevertheless, systems such as this are considered “non-agents” even though they may sometimes be treated otherwise. eg. shouting at a computer when it does not do what we want it to.

Only complex systems can be agents but a complex system need not be an agent. Whether an observed complex system is an agent or a non-agent cannot be known for certain and different people will attribute agency, or “mind”, to different things. For instance animals at various stages in the evolutionary tree.

Although our attribution of mind to an entity affects how we treat it or view it, and perhaps how it responds to us, in terms of describing, predicting or explaining observed behaviour, it makes no difference at all. An alien investigator could examine human society and come up with detailed models with considerable predictive power under the assumption that humans are automata. The unpredictability of individual actions could be taken as simply due to complexity rather than agency.

Knowledge from authority

As well as making observations and trying experiments of our own, we necessarily learn from the observations and experiments and mistakes of others. There is not enough time to do it for ourselves and this greatly increases what we are able to know. This raises the question of which of these authorities we can trust, especially if there is disagreement between them. The modelling of the environment moves from coming to conclusions about what objects do to how people behave and whether they can be trusted as authorities. To work out the reliability of the authorities themselves and whether to take on board, either provisionally or with conviction, their findings.

In these ways we can build up ever more complex, predictive and accurate models of our environment. Our empirical knowledge grows.

1. Bloom 2005: Paul Bloom, Descartes Baby, 2005
2. Elizabeth Spelke, “Initial Knowledge: six suggestions”, Cognition, 50 (1994), 431-445