Barnacle Geese, Bright Lines and The Regulation of Beliefs

Reasonable People #7: challenges to the idea of consistency as a general principle, and why other people can be a foundation for a system of reasons.

What’s good for the goose

There’s a story about the Barnacle Goose which is hard to believe. People on the Atlantic coasts of medieval Britain believed these birds to be fish, not birds. Nobody saw their eggs, or their young, or saw them sleeping in their nests. Instead they saw the geese diving under the water, and, the reports go, supposed that sometimes they simply declined to surface and spent the night down there. Barnacle geese were thought to be born from barnacles on driftwood, or (in some versions) from fruit from “goose trees”.

There are stories of this dating from at least the 11th century. In the 14th century Sir John Mandeville’s Travels reports:

I told them that in our country were trees that bear a fruit that become birds flying, and those that fell in the water live, and they that fall on the earth die anon, and they be right good to man's meat. And hereof had they as great marvel, that some of them trowed it were an impossible thing to be.

How hard is consistency? Two challenges

Last time (RP#6), I reviewed Nick Chater’s book The Mind Is Flat which makes a case for our beliefs and behaviours being radically inconsistent, lacking the stability of hidden representations in our mental depths, and instead being continually improvised and adjusted to fit what were perceive in the moment. I argued that Chater missed an important source of stability in our beliefs and preference. This could come from, I said, from agreeing with our publicly shared statements. Rather than mental depth, consistency-from-without.

A reader, K., wrote to say that my argument doesn’t make sense:

There seems to be some circularity in your argument, perhaps even a double circularity.

1. We act is if self-consistency is important, and therefore rationalize what we do post-hoc. Yet ‘consistency’ in behaviour and choices is a human construct. That means we are at liberty to stick that label on any set of behaviours we like. Describing ourselves as consistent is no more than applying that term to behaviour—like Humpty Dumpty, the word means what we want it to mean.

2. The same applies at the meta-level: the broad behavioural patterns to which you allude as evidence of consistency, our apparent consistent tendency to seek to demonstrate how consistent we are... that too is just a set of behaviours that we call consistent. It neither proves nor disproves that we are consistent.

Let me try putting these two challenges in my own words:

The first challenge is that it isn't a given that self-consistency is a priority. What grounds are there for assuming that people hold a common recognition of this as a goal? And if it isn't a common principle, then the bootstrapped consistency from without argument unravels.

The second challenge is that consistency is just a label, rather than an objective fact about our behaviours. If I am free to label any two actions as consistent with each other (but you might label differently) it doesn't prove anything about actual consistency (whatever that is), nor does it allow a foundation for us to agree that I am being consistent.

Both of these are serious challenges, and addressing them helps uncover something important about the nature of reason. The Barnacle Goose is going to help.

One: Do we value consistency, and why?

For the first challenge, I think I can see two responses: theoretical and empirical. Theoretically, we could imagine various reasons why we would value consistency. Even if just for reasons of parsimony, being consistent is desirable. It is the foundation for any system of rules - if you don't keep the same rules then you don't have a system. Consistency minimises your cognitive effort, just as habits of behaviour do (freeing you from extra decision making), and for the same reason that lying is harder than telling the truth (you borrow the consistency of what actually happened, rather than needing to make stuff up). It also allows you to predict what someone else will do, which seems to make it desirable in others too. True, "a foolish consistency is the hobgoblin of little minds", but this sentiment only stands in contrast to the wisdom of consistency as a general rule. Finally, physical reality is mostly consistent, and given our success at inhabiting the environment (and sense of betrayal when the ice breaks, the weather changes unexpectedly or whatever), it isn't surprising what we also want something of that constancy from our fellow humans.

Those are some theoretical arguments. Empirically, is consistency/coherence honoured where arguments are exchanged? It seems to be widely taken for granted that inconsistency harms credibility, and despite the common accusation that psychology is a science of the obvious I didn't have a lot of luck searching for direct tests of the idea. Certainly the wild inconsistency of some politicians, who not only survive but seem to thrive, suggests that inconsistency isn't fatal for credibility.

If anyone does know where I can find evidence that inconsistency harms your perceived credibility or trustworthiness, please let me know. (Like a lot of brief literature searches I'm left with the feeling that there could be a whole cottage industry of studies out there which would address this topic, but I just didn't have the right keywords or wherewithal to find it).

Conclusion: consistency is widely and highly valued, but it is interesting to explore when consistency is over-ruled, or inconsistency praised.

The story of the Barnacle Goose looks like a simple failure of consistency, but the belief that these birds were fish had a motivation. If the geese weren’t birds then they could be eaten during Lent fasting days, when meat of flesh or fowl was forbidden.

The belief about the Barnacle Goose creates an inconsistency (birds are flesh, except the Barnacle Goose which is a fish). But it also resolves an inconsistency (between “we should be fasting in Lent” and “we are going to eat meat”).

We can wonder why people in the middle ages didn’t just say one thing in church, and do another in the kitchen (as we know people so often do in the bedroom). The answer takes us to K.’s second challenge.

Two: Can you relabel any inconsistency as consistent according to a different rule?

The second challenge is harder to rebut, because it isn't wrong. Labelling of behaviour as consistent is a challenge for two people to agree on. Let me try and explain why. Suppose I come home on Monday with three apples and we share them out 2 for me, 1 for you. That looks inconsistent with fairness, but maybe I explain that since I got them I should get the larger share. On Tuesday you come home with 3 pears, and I argue that we should share them out, but again 2 for me, 1 for you. That looks inconsistent with Monday’s rule (you would certainly feel so), but maybe I can explain that today I cleaned the whole house, which is harder than fruit shopping, so Monday and Tuesday are consistent with the principle "bigger share to the hardest worker", or maybe I say that apples are 10 times better than pears, so really you are paying me back for the apple I shared yesterday instead of eating all three myself. For any apparently inconsistent set of facts I could invent other principles or circumstances which might apply and make them consistent.

The problem is that individual statements, like individual facts, don't speak for themselves but derive their meaning from context. There's flexibility in exactly what context is relevant, so flexibility in the meaning of individual statements. What appear at first as inconsistency or exceptions can be explained away, and that explaining away becomes the subject of a secondary meta-level discussion of consistency in judgements of consistency.

The original accusation was that relying on consistency has a circularity. In a sense it does, in that there is no absolute foundation. But I prefer to think that the fact that we can, in general agree, on what is consistent and when exceptions are legitimate, shows that we are somehow able to bootstrap our common understanding on a sense for consistency, can calibrate our sense of consistency on our common understanding.

As with the first issue, it may be that parsimony does a lot of work here. Any mutual understanding of a set of rules on which we agree risks becoming infinitely complex if each of us has diverging understanding and diverging understanding on top of that of what constitutes divergence. Consistency between our understandings, as well as among ourselves, represents a fixed point which collapses the multiplying ambiguities.

Keeping exceptions in the box

So my argument is that, yes, for any set of beliefs or facts you could arbitrarily invent rules which turn inconsistency into consistency, but parsimony allows us to find some agreement between us on what - for the most part - counts as consistent. This is why it was important that the Barnacle Goose was reclassified as a fish. The reclassification allows flesh to be eaten, actually, while still honouring the general idea of fasting at Lent and the whole doctrinal superstructure behind it.

The exception shows us something important about what would happen if you just allowed a blatant contradiction. This means jettisoning either an entire belief system (in this case Christianity) or the whole principle of consistency. Even something as trivial as choice of lunch threatens an epistemic explosion which could touch everything you believe about the world. It isn’t that consistency is self-evident: instead, it is arbitrary inconsistencies which are too easy to justify.

When we must break a rule we struggle to maintain the rule by containing our rule breaking within an exception. What stops the endless multiplication of exceptions? Potentially not much, which means we have to be very careful about the rules, and when exceptions are allowed.

George Ainslie writes about this in Breakdown of Will. Addiction is easy to slip into, he says, because the general pattern is not set by single actions. No one drink makes you an alcoholic, so we can always find an exception which allows one more drink while denying that we’re alcoholics. As with drinks, so with rule breaking in general. You have to be careful about starting down that primrose path.

Part of the counter-measure to this proliferation of exceptions is what legal theorists call Bright Lines - clear rules with little ambiguity. “No unnecessary drinks” or “Stop drinking when I’m not enjoying it” are not bright lines. “No drinks”, or “No drinks before 5pm” are.

Even when we break a bright line, we still might try and limit the rule-destroying power of exceptions. Eating the Barnacle Goose breaks Lent fasting rules, but the invented exception is a bright line: a single species is allowed, allowing the Lent fasting rule to still be honoured in the most part.

In the middle ages it seems this “bright exception” created an uneasy truce between principles. Irish bishops condoned eating the goose, but we only know this because they were criticised in turn by other medieval religious scholars for “leading people into sin”. In 1215, at a council in Rome which including over 400 bishops, the Pope issued a corrective condemning eating Barnacle Goose in Lent and affirming that they were birds regardless of their manner of birth.

Even in pious Ireland this decree was widely ignored, which I like to think is testimony to the fine balance the Barnacle Goose case occupied at the intersection point of different beliefs and desires of the medieval mind.

(Gerard’s Herball, 1597)

Choice Feedback

Moving away from Geese, but still on the topic of The Mind is Flat and the “consistency from without” argument, the Choice Blindness Lab got in touch to share work they’d be doing with Chater and others which anticipated my argument (and which shows that maybe my argument “against” Chater wouldn’t be something he’d necessarily disagree with).

Johansson, L., Hall, L., & Chater, N. (2011). Preference change through choice. In R. Dolan & T. Sharot (Eds.) (2011). Neuroscience of Preference and Choice. Elsevier Academic Press. pp. 121-141.

Which has a great opening paragraph

In an unforgettable performance by the English comedian Eddie Izzard, he portrays the Spanish Inquisition as conducted by the wimpy Anglican Church. Playing a supremely domesticated inquisitor, he offers the accused heretics the intriguing choice between “Cake or Death?” Apart from the hilariousness of this scene, the social scientist can appreciate it as one of the last safe havens for neoclassical economics, rational choice, and expected utility theory, because people really and truly have stable and identifiable preferences to help them decide between cake and death. There is simply no amount of social-psychology shenanigans that could push this preference around (no matter how little Festinger would pay people for choosing the death sentence, it would not generate enough cognitive dissonance to sway anyone) (Festinger, 1957).

And presents a framework for studying the feedback effects of choice - how we are influenced by what we have previously stated or shown we want, and how this can create a kind of “self-herding” (Ariely & Norton, 2008) where choices initially influenced by arbitrary factors become locked in by the pursuit of self-consistency

The most troublesome thing about the dichotomy between stated and revealed preferences is not the fuss over whether actions speak louder than words (they do), but rather that talking is one of the most ubiquitous actions we engage in, and therefore one of the most obvious channels for the study of choice feedback effects. Thus, stated preferences are always prime and potent candidates to change and induce further preferences, particularly as these statements very seldom are allowed to stand undisputed in the social fray of our lives….Here, we find both a glimmer of rationality in the distribution of information traveling between minds – in the asking, judging, revising, and clarifying of critical, communal discourse. (p125)

References from this chapter:

Ariely, D., & Norton, M. I. (2008). How actions create – not just reveal – preferences. Trends in Cognitive Sciences, 12(1), 13–16. doi: 10.1016/j.tics.2007.10.008.

Ariely, D., Loewenstein, G., & Prelec, D. (2003). “Coherent arbitrariness”: Stable demand curves without stable preferences. The Quarterly Journal of Economics, 118(1), 73–105. doi: 10.1162/00335530360535153

In this recent short commentary, they pursue the idea that the feedback effects of choice don’t just work on solitary individuals, but populate the representational space between agents:

Pärnamets, P., Johansson, P., & Hall, L. (2019, September 1). Letting rationalizations out of the box. https://doi.org/10.31234/osf.io/q96g7

Thanks

Thanks to those who got in touch, and to J. who I discussed the ideas here with (fwiw, J. thinks the principle of consistency doesn’t need more support than our grounding in physical reality, making most of my argument here superfluous). And special thanks to E., whose off-hand mention of Barnacle Geese on Sunday evening got me here.

I’ll be back next week with less self-reflection and more links to research on reason, belief and persuasion. Get in touch if there’s anything you’d like to share, and please pass on this newsletter to anyone who you think might enjoy it.

Stuff

Book review: Sway by Pragya Agarwal, reviewed by Steven Poole in the Guardian. Sub: “Does ‘nudging’ work? And how useful is it to assume that people ‘are not naturally rational’? This is an urgent study of the political harm of bias”

Link: A contemporary example of the use of bright lines in reasoning: Rules, Law and the Virus from Spinning Hugo

…And finally

This classic from Izzard: