Tuesday, May 5, 2009

5.2.6 Wesley Salmon and the Problem of Induction

Wesley Salmon was a 20th century philosopher who took up the study of how causality functions and why the transfer of information from one spatio-temporal location to another serves as its fundamental mechanism. He continued the investigation into questions of inference, induction, and the scientific process.

According to Salmon, we all believe we have knowledge of facts that extend beyond what we directly perceive. Our view of events is severely limited by both space and time. Based on our limited experiences we presume to predict future events. Take this hypothetical situation: suppose that you have drawn a number of balls from an urn and discovered them to all be black in color. You might infer, therefore, that all the balls from the urn will be black. This is an "ampliative" inference: the conclusion asserts something that is not found in the premises – nothing in the past record of choosing only black balls implies anything about the color of balls you will draw in the future. Some black balls have been drawn from the urn - therefore can one conclude that they will be black as we continue drawing? No, not at all, because in this sort of “non-demonstrative” logic, it is perfectly possible for the conclusion (all future balls will be black) to be false even if the premises are true (all the balls so far were black). Although this example is contrived, that is exactly the kind of judgment we make about the prospects for a sunrise each morning.

A related form of this scenario is called the “Black Swan” problem – it may be that all swans you have ever seen before are white, but you would not be on firm logical ground to conclude that all swans in the world are white. Even a single black swan (which was finally seen by Europeans when they first visited Australia) would invalidate your premature conclusion.

Hume and many others have noted that such inferences would be valid if we could have recourse to some sort of principle of uniformity in nature. If we could prove that the course of nature is uniform, that the future will be like the past, then we would be justified in generalizing from past cases to future cases - from the observed to the unobserved. We have found by experience that nature has exhibited a high degree of uniformity and regularity so far, and we infer inductively that this will continue. Even though we all have many beliefs about unobserved worlds, and in some of them we place great confidence, they are without a solid rational justification. By habit, we assume an intrinsic uniformity in the processes of nature. Such a belief seems easy to rationalize - these inferential methods of both common sense and science have proved themselves by their results. No other method can claim a comparable record of successful accomplishment, and probably most importantly, there is no compelling evidence or reason to believe differently. Consider the amazing technological and scientific discoveries that have been made using these techniques. However, it is easily shown that this is a circular argument, and once again we arrive at the necessity of making an assumption – the world is regular and, within limits, it conforms to predictable and regular patterns. Other theories may explain our experiences here in the world - the theory that the sun has risen every day so far but will stop doing so tomorrow fits the existing data as much as the theory that the sun will always rise. But is it just as believable? The answer is no - it is not. It is clearly an ad hoc explanation designed merely to fit the data, but can offer no explanation or predictions that are of any use. Given the two options: uniformity vs uniformity up to now, followed by chaos tomorrow, the former is far more deserving of belief than the latter.

The universality, uniformity, and predictability of nature are concepts that make their appearances in all debates involving inference and induction. It seems that for inference from past experience to be a valid way of reasoning about future events or current, but unseen events, we must believe that the future will be like the past. This implies a uniformity that spans time and space. But is this a good assumption? If we allow ourselves to be trapped in the circular explanation, it is not good enough.

But why should we limit ourselves in this way? Isn't predicting the sunrise substantially different than predicting whether the next ball will be black or not? We know almost nothing about the unseen balls in the urn, but we know an enormous amount about how the sun and earth are related. Thanks to Newton, we understand orbital mechanics. We can trace the history and describe the structure of the solar system, and the laws of gravity and angular momentum are well-known. Even more, we can apply our knowledge of the Milankovitch cycles that describe the 41,000 year cycle of Earth's precession, and even bring in General Relativity to describe the subtle changes in the precession of Mercury's perihelion. All these facts and theories support the modest contention that tomorrow the sun will rise. Combining the findings from the sciences and many other disciplines to form a network of supporting evidence is the essense of "Coherentism", discussed in the next section. It is a hopeful alternative to the tautology we are trapped in when we use induction to prove induction. The next section describes this concept.