Another installment of the in-progress paper.

Draft only. Comments welcome

3.2 Complementation

A second theme in understanding how a tool such as Rationale can make us smarter is complementation: the tool complements our minds’ natural strengths and weaknesses.

Our reasoning abilities are a function of our basic cognitive capacities, which depend in turn on our brains. Those brains are the result of a long, accidental, accretive and incomplete process of evolution. They were “designed” not for the sublime arts of logic and disputation but to enable us to survive and propagate in the physical and social environments of our ancestors going back long before the emergence of homo sapiens. Our reasoning abilities are a late acquisition, due not to any newly-crafted reasoning module but rather to our learning how to exploit in a new way cognitive capacities which had already been wired in for other purposes. That pre-given set of capacities includes some remarkably powerful and useful functions, but does not provide everything that might be needed for flawless, general-purpose reasoning.

In light of this, a sensible strategy for improving human reasoning is to provide tools which, on one hand, provide ways to bypass or make up for the deficiencies or limitations of our innate capacities, while on the other, taking advantage of their distinctive strengths. Rationale makes use of both these strategies.

The most important respect in which a tool like Rationale compensates for inherent cognitive limitations is by augmenting our short-term memory (STM). The need for such augmentation can be summarised in three points. First, reasoning and argumentation can become very complex. Second, as we standardly do things, that complexity places huge obligations on memories. However, third, our innate STM is quite limited – much more so than people usually realise.

The standard cliche about human STM is that it can hold “7+/-2” items. Thus 8 digit phone numbers are quite a bit harder to remember than 7 digit ones. The 7+/-2 figure originated in famous research in the 1950s (Miller ref). However the figure should be treated with caution. The original research concerned our capacity to recall random sequences of meaningless items. Recent research suggests that for such sequences, Miller’s figure may be an overestimate. At the same time, STM can be increased when items are “chunked” or meaningfully grouped. And with intense training, people are able to perform remarkable feats of short-term memorization, such as remembering dozens of random digits.

Nevertheless, it is clear that, in normal circumstances, human STM has quite a low ceiling. This can be easily illustrated by asking two people to play tic-tac-toe without using pen and paper or any equivalent – i.e., to record moves in their “mind’s eye” rather than on a board. Most people find that completing the game is a fun challenge, in part because it is hard enough just keeping track of the state of the game, let alone making moves.

Meanwhile it is obvious that arguments can be very complex – far more so than a mere game of tic-tac-toe. For example the case put forward by Jim Garrison in the trial scene of Oliver Stone’s movie JFK consists of dozens of pieces of evidence woven into an intricate web. Even this case (let alone the galaxies of arguments and responses in the larger “Who killed JFK?” debate) is more complex than can be held, organised and evaluated purely in the head by anyone other than, perhaps, an idiot savant.

This point holds true even in the mundane territory of our everyday disputes or academic altercations. The issues and arguments are often if not always larger than our unaided minds alone can easily embrace; and retaining even some of that material is effortful and prone to loss, confusion and confabulation.

Yet we do presume to think through such complex cases, and so by practical necessity we make use of external aides de memoir; for example, recording and organising our thoughts in notes, essays or books – i.e., in some form or other of prose. Similarly, a central function of an argument mapping program such as Rationale is to function as an external “memory”, or memory-extension, for reasoning. Maintaining stable structured representions of arguments or debates of effectively unlimited complexity is trivial for an appropriately-programmed computer. Thus, when using such a package to help us think our way through a set of arguments, we are taking advantage of a great strength of computers to compensate for an inherent weakness in our own capacities.

For external representations of reasoning to be useful in our thinking, they must be such that we can interact with them fluidly. On one hand, we must be able to create and modify the representations easily; this was touched upon above in the discussion of usability. On the other, we need high-bandwidth access to the information contained in those representations. Ideally, we’d be able to “read” the representations faster than we could think about what they are representing. Here, argument maps are designed to exploit some remarkable strengths of our standard cognitive equipment. I claimed above that an argument mapping software package can be more usable as a tool for reasoning because they exploit representational resources which are neglected by typical argumentative prose – resources such as colour, shape, line, and position in space.

The use of these resources is a great advantage because our “hard wired” mechanisms for visual cognition are designed to process information coded in these basic dimensions with extraordinary efficiency. When you look outside the window and see a tree, your perceptual system accepts and processes a vast amount of basic visual information and reliably delivers a correct high-level interpretation in a fraction of a second and with no perceptible effort on your part. This is an example of what psychologists often call “pre-attentive processing” – information being taken up and utilised so fast that you didn’t even have time to shift your attention to it.

Colour, shape, line, and position in space are all aspects of a scene which are, or can be, pre-attentively processed. Any tool for reasoning which wants to optimize the transfer of information from external representations to central cognitive processes ought to exploit pre-attentive processing as much as it is effectively able. Argument mapping software does exactly this, with the result that representations of complex reasoning can be accessed, and hence utilized in thinking, much more quickly and easily than in standard prose formats.

The most obvious example of this is the use of colour to code for “polarity,” i.e., whether one proposition is (taken to be) supporting or opposing another. In standard argumentative prose, polarity is something which must be “computed” through slow, effortful and error-prone high-level interpretative process. Consider for example this passage:

Like us, dogs, wolves, chimps, and macaques are all social animals. They show that we are not unique inventors of empathy and morality.

It contains two propositions. Assuming that together they constitute a simple argument, which one is evidence in relation to the other? And is it supporting or opposing? Answering these questions requires a bit of careful attention. We must understand the sentences, and think about how they relate to each other. If however the argument is presented in a standard argument mapping format:

then, assuming you are well-versed in the relevant conventions, you will pre-attentively process the green on the lower claim, and know that the lower claim is being presented as a reason for the upper one, even before you’ve actually read and understood those claims. In other words information about evidential structure is being conveyed in manner which is extremely fast, requires almost no effort, and unambiguous – much like seeing the tree outside the window.
This example focused on the use of colour to indicate polarity, but shape, line and position in space are working here in much the same way. Indeed, the visual design exploits all four of these dimensions simultaneously, with each reinforcing the message conveyed by others.

Thus, an argument mapping software package is exploiting an impressive strength of the human mind, namely its ability to process and interpret certain kinds of basic visual information very rapidly, effortlessly and reliably.

More profoundly, the greatest strengths of the human mind are its abilities to comprehend natural language and evidential relationships. Despite the valiant efforts of computer scientists over many decades, computers still lack anything seriously resembling human intelligence (notwithstanding world-champion chess programs and the like, which are at best electronic idiot-savants). Thus a software package designed to improve human reasoning must still rely on our minds to do core “heavy lifting” involved in the performing and evaluating reasoning.