Monday, December 12, 2011

Analogical reasoning

Analogies represent an important yet confusing domain of reasoning. There is no intuitive simple way to understand information processing related to analogies. Mathematics and computer programming do not deal with analogy. It is extremely difficult to program computers to recognize and process analogies. This also means that it is very difficult to theorize the processes.

Yet, we know that managers need to constantly engage in analogical reasoning. Entrepreneurs use analogies to evalute, examine, and sell novel ideas (e.g. Cornelissen & Clarke, 2010 in AMR). The whole Harvard case method popular in almost all business schools is based on the assumed ability of analogies to prepare students to manage firms and to solve problems.

There is a very interesting forthcoming laboratory study on analogical reasoning to be published in Strategic Management Journal by Lovallo, Clarke & Camerer.

The lab experiment on analogical reasoning

To examine the processes of analogical reasoning in managerial judgments, the authors asked experts to judge the expected returns for a sample of new ventures. They inially instructed the participants to take an "insider view" without analogical comparisons to other similar cases:
Please describe the path along which you see the Project proceeding. Start from where the Project is now and construct the most probable future scenario for the Project. Please create a timeline that describes the key steps, milestones, and actions that need to be taken to reach the Project’s goal, using as much space as you need (This should take about 15–20 minutes). After you have finished, please answer the questions on the following pages.
 They then asked the participants to use analogies to re-examine their estimates:
What two categories of investments or potential investments are most similar to the Project (e.g., founder-seller, early-stage, technical-risk, public company)? You can define/create whatever categories you think are the most relevant to the Project.
The authors found that the use of analogies led 82% of participants to lower their estimates (initial estimates were far greater than industry averages).This, I thought, was not particularly interesting as it can be simply an example of anchoring bias (in this case, a useful one). The interesting observation is this:
One finding from the study is that people do not seem naturally inclined to form a broad reference class of projects even when encouraged to do so. [...] the vast majority of reference projects were described as successes.These results are disturbing, as they suggest that reference class forecasting is itself open to bias in the recollection of reference projects (Kahneman and Tversky, 1979).
Analogical reasoning should not be limited to successes

It is clearly alarming if managers mostly base their analogical reasoning on success cases. People justifiedly complain about From Good to Great for its lack of attention to failed firms.If we compare situations we encounter to success cases, we will become overly optimistic about the characteristics of our situation that match successes while at the same time ignoring characteristics that would match failure cases. People love to read and hear good stories. Also, those who succeed are more willing to share their stories than those that fail. Journalists know this and tend to write about successful innovators and turn-arounds. Business professors know this and tend to write and teach case studies of the heroic success stories.

To counter these tendencies, we might need to instruct managers to engage in systematic analogical reasoning. When thinking about an investment or other decisions, managers should see that their discussions and private reflections consider analogies not only to similar success cases but also to an equal number of failure cases. Lovallo et al. forthcoming paper leads to a similar conclusion, although they propose a much more systematic methodology.

Two possible theories of analogical reasoning

I have yet to read a definite paper on the philosophy of analogical reasoning. Personally, I think there are two ways to go about it. First, we can assume analogies to be very complex and follow some form of parallel processing where numerous attributes are matched and conclusions are drawn from a complex body of tacit knowledge. The parallel processing can be assumed to be so complex that it is practically untheorizable on micro-level. The best we may be able to do is to see broad tendencies of  past experience or working memory recall to influence outcomes.

Second, we can assume analogical reasoning to follow working theories that are commonly tacit but can be made explicit. That is, analogical reasoning may repreresent a systematic comparison of characteristics and the application of knowledge concerning the relationship of matching characterics on outcomes of interest. We may have never heard of a "snow lion" (an imaginary animal), but through analogy we can figure out the likely characteristics: it would be analogous to snow leopard in being white and furry. It would likely eat other relatively large mammals such as goat, analogously to normal lions. These would be predicted by the knowledge-based theory of categorization -- traits that we know to have either functionality in snow-related environment or likely to be inherited across sub-species.The analogical reasoning is not pattern-matching, but the application of causal knowledge to estimate likely similarities.

The first option mystifies analogical reasoning as something that arises from the complexities of the human brain. The second option suggests that analogical reasoning is really just the application of existing knowledge to match premises to likely outcomes - not qualitatively different from more formal reasoning tasks, except that the initial premises attended to arise from specification of one or more analogous exemplars. The second option is the only account of analogical reasoning that allows discursive consideration of analogies. Classroom discussions of case studies is not about matching patterns, but about illustrating and memorizing knowledge concerning causal relationships.


Analogy and metaphor are difficult topics that can easily become mystified. They are linked to powerful and complicated processes of reasoning that we need to understand better. The forthcoming paper by Lovallo et al. is a nice example of work trying to make analogical reasoning more explicit with simplicity and clarity.

I'll conclude with my favorite passage of James Joyce, an application of analogy (metaphor) to motivate selective reasoning about the properties of women.
What special affinities appeared to him to exist between the moon and woman?

Her antiquity in preceeding and surviving successive tellurian generations: her nocturnal predominance: her satellitic dependence: her luminary reflection: her constancy under all her phases, rising and setting by her appointed times, waxing and waning: the forced invariability of her aspect: her indeterminate response to inaffirmative interrogation: her potency over effluent and refluent waters: her power to enamour, to mortify, to invest with beauty, to render insane, to incite to and aid delinquency: the tranquil inscrutability of her visage: the terribility of her isolated dominant implacable resplendent propinquity: her omens of tempest and calm: the stimulation of her light, her motion, and her presence: the admonition of her craters, her arid seas, her silence: her splendour, when visible: her attraction, when invisible.

James Joyce, Ulysses Vol. 2, p. 110