Strategic ambiguity or unknown unknowns?
Part 4 of the 6-part series, Modeling for metascientists (and other interesting people).
Many social scientists do not consider formal models a useful tool in theory construction. They might defend this position by arguing that “formal models make assumptions that are too simplistic, too unrealistic, and too arbitrary.” Instead, they rely on verbal models.
In turn, formal modelers complain about the ambiguities associated with verbal models, including imprecise constructs, implicit assumptions, and predictions based on intuitive reasoning rather than deduction.
In disciplines where verbal models dominate, theoretical ambiguity is widespread. This raises some uncomfortable questions: Don’t scientists notice these ambiguities? And if they do notice, could it be scientists actually prefer ambiguity?
Imprecision can be useful. Daniel Dennett, for example, talks about the ‘heartbreak of premature definition’. It’s unwise to settle on definitions before we know which elements to include and which to exclude. Likewise, perhaps researchers shouldn’t commit to a particular set of assumptions too soon; perhaps the imprecision of verbal models, at least in the early stages of theory construction, is a feature rather than a bug.
This argument seems defensible. However, people who do think that it is better to be “vaguely right, than precisely wrong” — as John Maynard Keynes put it — should acknowledge their pragmatic use of imprecision.
But this rarely happens. Instead, verbal theories are usually depicted as solid and precise. Why might this be?
First, it is possible that researchers sincerely regard their verbal models as solid and precise. They just don’t see the ambiguities.
This might explain most cases. For example, we all know that we will die one day, but few of us know precisely when. Drawing from Donald Rumsfeld’s epistemic taxonomy, our expiration date is a “known unknown” — we know exactly what we do not know.
But there is also information that we don’t know that we don’t know. Verbal models make it easy for such “unknown unknowns” to be concealed, even from researchers themselves. One of the merits of formal models is to shine a light on the darkest corners of verbal models, making all assumptions and reasoning explicit. The “unknown unknowns” are thus transformed into “known unknowns”.
The second possibility is more sinister.
Incentives might favor researchers who strategically, though not necessarily intentionally, use ambiguity. Scientists are motivated not only to discover truth, but also to have impact. One way of achieving impact is to propose inspiring ideas that can be interpreted in diverse ways. An ambiguous verbal model is better for this because it can more easily accommodate different ideas and findings.
As Eisenberg (1984) noted, “Ambiguity is used strategically to foster agreement on abstractions without limiting specific interpretations. … Strategic ambiguity must be viewed as a continuum, from most clear to most ambiguous; the more ambiguous the communication, the easier it is to deny specific interpretations. … The more ambiguous the message, the greater the room for projection. When an individual projects, he or she fills in the meaning of a message in a way which is consistent with his or her own beliefs.”
If motivated more by impact than truth, researchers may prefer the ambiguity of verbal models over the precision of formal ones.
Again, strategic ambiguity does not imply intentional ambiguity. People are not always aware of their motivations and goals. Incentives can shape selection processes to favor behaviors that lead to preferential transmission of ambiguous, verbal models over precise, formal models, without any individual awareness of these processes.
For instance, a scientist might rationalize a dislike of formal models for the simplifications it demands, without realizing that this dislike is influenced by the concern that formal models will reduce the impact of their verbal models. As David Hume observed, “Reason is the slave of the passions”. Alternatively, researchers may well be aware of their goals, but not of the extent to which such goals end up shaping their beliefs.
What is to be done?
We should strive for transparency. If you use ambiguity deliberately — for example, to avoid premature theorizing — signal your intentions so that readers know what you are doing and why. If, instead, you believe that your theory is solid and precise, welcome the careful scrutiny that only formal modeling can provide.
Above all: Don’t be ambiguous about whether you are being ambiguous!
Want to learn more about the merits of modeling?
Frankenhuis, W. E., Panchanathan, K., & Barrett, H. C. (2013). Bridging developmental systems theory and evolutionary psychology using dynamic optimization. Developmental Science, 16, 584–598. https://doi.org/10.1111/desc.12053
Smaldino, P. E. (2020). How to translate a verbal theory into a formal model. Social Psychology, 51, 207–218. https://doi.org/10.1027/1864-9335/a000425