When scientists attempt to explain observations of behaviour in humans and other animals, they often use language that evolved for informal discourse among people engaged in every day social interaction, like this:
Similar comments can be made about the terminology used in many philosophical discussions about minds, cognition, language, and the relationships between evolution and learning.Often the theories put forward merely list components and processes characterised using ill-defined labels adapted from common sense, or else list information contents allegedly used (e.g. beliefs, concepts, theories), but do not attempt to specify the explanations in terms of a design for a working system.
It's as if someone tried to explain how a car engine works by listing and labelling parts without indicating how any of them work or how they interact.
We need to understand that in talking about a mind (or a major subsystem of a mind) we are talking about a complex system with many concurrently active parts -- that work together more or less harmoniously most of the time but can sometimes come into conflict.
These parts are organised in an information-processing architecture that maps onto brain mechanisms in complex, indirect ways that are not well understood.
So, when studying some human (or animal) psychological capability or limitation, we should ask questions like this if we wish to do deep science:
If we think of features of humans and other animals such as consciousness, intelligence, attention, memory, emotions, autonomy in this 'design-based' way (adopting what John McCarthy now calls 'the designer stance') the sorts of questions we can ask and the sorts of theories we can consider are expanded in an important way.A design for a working system (microbe, ant, chimpanzee, human, robot) will specify a complex virtual machine with many coexisting, interacting information-processing components (as explained in this PDF presentation).
Since there are many components, it is possible to consider different designs for working system that use different combinations of such components, and different versions of the components. This sort of variation in designs is evident in the products of biological evolution.
A corollary is that where we are naturally inclined to think of a binary division such as a division between animals that do and do not have some feature X (consciousness, creativity, autonomy, emotions, planning capability, free-will, etc.) the design based approach replaces the binary division by something more like a taxonomy or a grammar, allowing for a (possibly large) collection of cases where there are typically many discontinuities
Example:
Many people (including me once) assume that there is a binary division between reactive and deliberative control mechanisms.After hearing several presentations and reading several documents making use of these labels in confusingly different ways, I eventually realised that people were interpreting the division in different ways because the space of possible designs had a kind of complexity that had not been studied properly and people were basing the distinction on different 'cracks' in the space.
For example, some people were using 'deliberative' to refer to any system that could, in some sense, evaluate alternative action possibilities and select one, whereas others used the label to refer to more complex systems that can plan more than one step ahead when taking decisions.
When I looked closely, I found that there were several more important sub-divisions between different sorts of deliberative competence, and documented them in this discussion note on 'fully deliberative' systems.
I don't claim that the analysis of that document is complete: there may be more sub-cases to distinguish.
When considering any X and asking which animals have X, how X evolved, what X's costs are, what Xs benefits are, which neural or other mechanisms are involved in X, etc. a good heuristic is to ask
And that generally has the result of replacing a binary divide between things that do and do not have X with a sometimes large collection of varieties of X, and a large collection of intermediate cases between not having a particular sort of X and having it.
- How many varieties of X are there?
- what sorts of distinct components, that might have evolved separately, are involved in different varieties of X?
This idea was used in an analysis of the notion of 'free will' originally posted to a 'usenet' news group, now available here.
The previous section pointed out a consequence of this study of varieties of architecture that might have been designed or might have evolved in response to explicit or implicit requirements such as the pressures of an ecological niche, namely that we usually need a richer ontology of types of design than can be expressed using binary distinctions normally assumed.Another consequence of the approach is that the process of designing a complex working architecture, testing it and finding problems that need to be addressed by improvements in the design, often teaches us that there is a much richer variety of possible internal states than we might have considered possible in advance.
Systems with complex virtual machines that include concurrently active interacting subsystems, including some subsystems that monitor and control others, can have a richer variety of internal states and processes than can be defined in terms of varieties of external behaviour, or even relations between inputs and outputs. For example, a system can run internally and have no connections with output signals most of the time, even though it occasionally is linked to inputs and outputs.
A simple example could be a complex virtual machine that is capable of playing many different games, and at any time practices some of those games internally by playing itself, e.g. at chess, or draghts (checkers), as a result of which its competence in those games increases, though there is no external sign of those changes unless it engages in a game with an external player, which may never actually happen.By studying the variety of internal states that the architectural design (the information-processing architecture) of some organism makes possible we may find that to understand and explain how the organism works we need a much richer ontology of states and procsses than would be suggested merely by watching its behaviours and trying to classify them.Its input and output channels may have capacity limits that limit the total number of games that the system actually plays in its lifetime, and that limit may be significantly lower than the number of different games it has the competence to play.
This is particularly rue of humans: there is no reason to suppose that the ontology expressed in our ordinary language concepts for talking about mental processes, or even the extensions to that ontology developed by psychologists and psychiatrists as a result of interacting with and experimenting on humans is rich enough to account for all the important phenomena of human life: instead we need a much richer ontology of states and processes derived from a good theory of how the system works. This is similar to the way our understanding of the variety of types of material substance had to be substantially revised when we discovered the underlying architecture of matter, as composed of atoms of various sorts that can combine to form molecules of various sorts that can be arranged in configurations of various sorts -- none of which was dreamt of prior to the development of modern physics and chemistry.
The issues raised here are pursued further in different ways in different online papers produced as part of the Cognition and Affect project and the CoSy Robot project, referenced below.A particularly relevant methodological discussion paper is
Two Notions Contrasted: 'Logical Geography' and 'Logical Topography' Variations on a theme by Gilbert Ryle: The logical topography of 'Logical Geography'.That paper discusses relationships between philosophy and science in the context of an attempt to clarify Ryle's notion of 'Logical Geography', showing that there is a deeper type of investigation, which I called the study of 'Logical Topography', which identifies aspects of some portion of reality that allow various possible kinds of concepts to be developed, in contrast with the study of the concepts that are actually in use, Ryle's 'Logical Geography'.The difference emerges in two ways: The study of logical geography assumes (a) that there is one collection of concepts whose relationships can be charted and (b) that this will answer philosophical questions definitively. The study of logical topography reveals (a) that the relevant aspect of reality can be divided up in different ways, leading to different logical geographies, and (b) that that reality may itself may have unnoticed complexity of structure, which, when explored in depth, shows possibilities that were not exposed by the original philosophical investigations.
On the basis of those ideas, the paper identifies a kind of philosophical theory building that has much in common with scientific theory-building (including the ability to introduce extensions to our ontology), and which uses abduction.
Originally installed here in 2005
Some of the ideas were
in
this paper
Prolegomena to a Theory of Communication and AffectIt distinguished various approaches to the study of mind: the design-based approach, the phenomena-based approach, and the semantics-based approach. I could have added more, e.g. the mechanism-based approach that starts from assumed 'known' mechanisms (e.g. neural nets) and tries to show how they can account for the observed, or introspected, phenomena. Daniel Dennett's "intentional stance", also referred to as "The knowledge level" by Herbert Simon and Allen Newell, could be mentioned here: it assumes that the individuals being studied are rational. I regard that assumption as both unjustified and unnecessary for studying and modelling biological information-processing systems (e.g. humans, ants, microbes).
In Ortony, A., Slack, J., and Stock, O. (Eds.) Communication from an Artificial Intelligence Perspective: Theoretical and Applied Issues.
Heidelberg, Germany: Springer, 1992, pp 229-260.
The design based approach is closely related to the study of logical
topography of sets of concepts, as opposed to the logical geography.
See
http://www.cs.bham.ac.uk/research/projects/cogaff/misc/logical-geography.html
Using a "design-based" approach is partly similar to adopting what Dennett has called the "design stance", in at least some interpretations of what he is saying, namely an interpretation in which one considers How something works and h How other things like it could be made.
There is a different interpretation of Dennett's design stance, summarised here
http://en.wikipedia.org/wiki/Intentional_stance
When we predict that a bird will fly when it flaps its wings, on the basis that wings are made for flying, we are taking the design stance. Likewise, we can understand the bimetallic strip as a particular type of thermometer, not concerning ourselves with the details of how this type of thermometer happens to work.On that interpretation one only assumes that something has been designed (possibly by evolution) to perform a certain sort of function and uses that assumption to predict what it will do. That does not use what I have called "the design-based approach", which involves trying to understand how it works. I suspect Dennett switched between the two interpretations.
The main difference is that what I have called the "design-based approach" emphasises the need for comparative investigations in different parts of both
I believe John McCarthy's use of the term "the designer stance" is
closely related to what I have called "the design-based approach".
See his paper
The Well Designed Child
Compare my
Kantian
Philosophy of Mathematics and Young Robots.