Wednesday, June 24, 2009

Is belief in the external world a hypothesis?

In the first three chapters of his book, The Problems of Philosophy, Bertrand Russell makes the case that we cannot know for certain whether there is actually a real world outside of and independent of our perception of it. Is belief in the external world a hypothesis?

Understand first that a hypothesis is a sort of educated guess. It is a conjecture, an idea, a reasonable suspicion, to be proven valid only by experience, and by experiment, experience’s substitute teacher. Yet even a valid hypothesis is not a certainty.

That a real world exists is a sound hypothesis, with much evidence to hand in support – but it remains a hypothesis. Russell argues that we cannot ever know or prove the independent existence of reality to a level of absolute certainty. I think he’s right, and I’ll go one better by saying that while we can know things well enough to suit our purposes, I don’t think we can ever know anything for certain.

Russell argues, for instance, that we can only know our own sense-data for certain. However, the existence of such phenomenon as “phantom limb syndrome”, wherein amputees feel distinct sensations from missing limbs, proves that even our own sense-data are suspect. I find Russell’s arguments shaky throughout, but due to constraints of time and space, I will try to address only a few select issues in this post.

Russell spends time discussing tables and the properties thereof, using them as an example of something real and commonplace that we take for granted. He argues that:

1. Our understanding of such a thing as a table comes not directly, but through sense data.
2. Sense data are not identical to the table itself.
3. What we know is therefore not the table in and of itself, but sense data
4. Sense data are unreliable and superficial representations of physical objects
5. Therefore the real table, and by extension, the real world is not truly known to us.

Although I agree that the existence of the external world is a hypothesis, Russell’s argument for this conclusion hinges on a common but fallacious assumption about knowledge, namely that knowledge is data. That the data we do have about a physical object may be incomplete or inaccurate does not negate our knowledge of that object.

The assumption that data is knowledge can be easily refuted, as anyone who has worked in the world of computers can attest. Computers can amass vast quantities of data and store it in databases. Is this knowledge? If it were, we would not have a separate word, “data”. But data is not completely separate from knowledge – it has the potential to inform knowledge.

The telephone company records every second that every cellphone customer spends connected to their network, and stores that carefully in company databases. Look at this data closely and you will discover that it’s a jumble of ones and zeros. Look at sense data closely – and science has – and you will discover it’s a jumble of bio-chemical and electro-chemical reactions. Is this knowledge?

Not yet – it has a long way still to go. First this data has to be organized into meaningful elements and components. It has to be put into a formation that is intelligible to the human mind. It is the job of the modern Information Technology (IT) department, and the computer programs they develop, to turn data into information – to put data “in formation” – so that it can be interpreted by a human being. Out of a mountain of data comes a series of reports, prepared by computer programs. Now the data has been turned into information. But is it knowledge yet? If mere data does not constitute knowledge on its own, does information?

Not yet – it still has a way to go. When I connect to the Internet, I can connect to billions of pages of information. The Internet is a galaxy of information, and I have access to all of it. But does having all the world’s information at my disposal, however well organized and well “in-formed” it is, mean I have all the world’s knowledge?

Of course not. Knowledge is more than data and more than information. If it were not, we would not have a separate word for it. But what is that secret sauce that turns a sandwich into a banquet?

When management at the phone company reviews the dozens of reports produced by the IT department, they don’t do so to admire the quality of the paper or the readability of the typeface or the logic of the categorization of the information. They review those reports in search of something actionable. They ignore information that can’t be used, however well “in-formed” it is, because information that can’t be used is by definition useless. And can we describe useless information as genuine knowledge? Of course not. For information to be regarded as knowledge, it must be useful for something; it must have a use-value. What good is information if you don’t know what to do with it? Knowledge is information that is usable. Knowledge is information with a use-value.

And here we can return to Russell’s arguments about sense data. Science has shown that our senses receive signals from the environment, some of which may indicate the presence of a physical object such as a table. Of course, our experience of the table is not immediate, but is mediated, meaning it comes to us through various media, among them light waves and sound waves, which tell us of the existence of the table.

Russell and many of his critics seem to agree that if our perception of the table is not immediate, it is invalid. Why this should be so is quite unclear, for surely we can perceive the table through various media, and can do so reasonably reliably.

This data is received by our sensory organs as a blooming, buzzing confusion. Our minds seek to arrange or formulate this data into some sort of meaningful, recognizable pattern. (This is why we sometimes see castles in passing clouds or a face on the moon.) This activity is analogous to the activities of the IT department at the phone company, when it turns mountains of data about cellphone use into meaningful information.

As Jeff Hawkins points out in his recent book, On Intelligence, our minds then engage in an extremely rapid process of comparing this sense-data to previously learned patterns, looking for something relevant. Our minds may be able to match patterns of incoming sense-data to previous experience, and if they do, the sense-data is translated into meaningful information.

This is a table, on the table is a cup and a pen and paper. There are two chairs, the floor is hard, and there’s a scent of flowers in the air, and the temperature is slightly cool. All this information is assembled from sense-data, given formation through pattern-matching and language, and then ever so briefly reviewed by the mind to determine how useful it is.

If on entering the room a colleague throws you a tennis ball, everything you have perceived, consciously or unconsciously, falls away, discarded by your brain as just so much useless information. Your brain focuses on the task at hand, catching or avoiding the throw. The sense-data that your mind translated into information about the ball has now become useful knowledge. Thus intelligence moves from data to information to knowledge.

No analogy is perfect and the analogy between our minds and computer and information processing technology is full of holes. But the parallels are sufficient to suggest that we may have created electronic information processing on the model of our own mental processing.

So what was missing in Russell’s account of the table? The use-value! How can we claim to any knowledge of such a thing as a table without understanding its use, its purpose, its telos? The moment we introduce the property of usefulness and purpose to a physical object, we possess the key to knowledge of it, because the concept of utility itself includes the concept of the relation of the physical object to humanity.

The very word “table” is a human construct, a symbol intended to convey the use of a physical object to other human beings. Its use as a word contains the implication that humanity has a relationship not only to tables, but to the physical world. This object before us is something that is reasonably flat and stable, so one can put a cup or dish on it without spilling, write a legible note on its surface, etc. It can have four legs, three legs, no legs, or it can be a stump or a flat rock. If we focus on data about the table – its color, how many legs it has, the material of which it is constructed, how its surface looks under a microscope, etc. – and ignore its use, as Russell and other writers appear to do, we will never really know or understand the table or the qualities of tableness.

There are numerous attributes of any particular table that are more or less irrelevant to its use-value as a table, and indeed to its status as a representative of the class of tables. These qualities fall mainly under the heading of “Useless Information”, and should therefore not be the yardstick by which we measure our knowledge of such a thing as a table. To do so, as it seems Russell and his critics have done, is to miss the point.

We don’t have all the data about a table or any other physical object, and indeed can never have it. Furthermore, because our perception of the table is not immediate, it is subject to any number of distortions internal and external to the perceiver. For these reasons and others, the existence of the external world remains a hypothesis. It is nevertheless a fairly sound hypothesis and a reliable guide to action – and that’s good enough.

One can only know the physical world insofar as one knows the use to which one seeks to put it. Our understanding of physical objects is inextricably linked to our intentions towards them, and the failure to recognize this can only result in the sorts of philosophical absurdities Russell warns us never to be frightened of…


Russell, Bertrand (1997). The Problems of Philosophy. New York, Oxford University Press.
Hawkins, Jeff (2004). On Intelligence. New York, Henry Holt and Company.