Loss aversion, one more victim to replication crisis? (no)

I saw somewhere a link to working paper by Gal & Rucker entitled The Loss of Loss Aversion: Will It Loom Larger Than Its Gain?, with the comment that loss aversion is one more psychological phenomenon not replicating. I was surprised by this claim because the mechanisms behind loss aversion, or something like it, are very much related to affect psychology. The context of replicability implies that loss aversion was never a thing, it was just a statistical fluke resulting from questionable research practices like social priming and Bem’s psi findings. But that’s not what the linked manuscript says. It’s a review that never actually questions whether such a phenomenon exists at all – rather it’s discussing the scope and an alternative conceptualization of the phenomenon. I’m not that familiar with this literature to assess whether the review really is impartial or whether it cherrypicks its findings (as it’s written quite obviously with a particular conclusion in mind, self-citing a lot), but clearly it should not be cited as evidence that loss aversion as a phenomenon is a result of QRPs. I’m a bit annoyed that the title and even the abstract plays like a clickbait and makes it easy to link the normal theoretical discussion about the limits of a phenomenon to the replicability issue.

In addition to enabling the misreading of this manuscript’s position in the literature, I was slightly miffed that it’s at least partly based on a fundamental misunderstanding of how the mind works. The authors describe a strong and a weak form of loss aversion in order to compare them to evidence, both of them in terms of it being a general, universal principle that can be applied to any human behavior: the strong form as absolute (that “one should not observe cases where gains have a propensity to be weighted more than losses of similar magnitude”, p.9), and the weak form as relative (“on average, one expects the data would largely reveal a greater impact of losses than of gains”, p.10). It may be that this is a feature in decision making research in general rather than a view held only by the authors, but from the point of view of affect psychology, it makes little sense. It’s a strawman, because I don’t think there is anything in psychology that can be considered a universal law like this, at the level of observable outcomes. Human mind does not work on “principles” or “laws” like this, because it is an immensely complex system of reacting, predicting, and self-correcting processes. There is no single process reaching through the whole of human mind, always (or even mostly, on average) producing the same results regardless of circumstances, because that would not be adaptive for the complex physical and social environment our mental machinery. And even if we focus on a very high level, it’s a dubious notion to begin with that all decision making would be governed by a single process translating all kinds of decisions into simple losses and gains.

I admit that I think some things as “principles” of human mind, and negativity bias (related but not identical to loss aversion) sounds like a good candidate, but it does not mean that at the level of observable outcomes, regardless of circumstances, we should see (absolutely or on average) a particular pattern of behavior. Rather, it means that some parts of the system tend to process information in certain ways, and in specific circumstances – where we can somehow control that specifically these processes are the ones influencing the outcomes the most – we can indeed see patterns in behavior.

That said, the alternative conceptualizations – such as propensity towards inaction or status quo – are interesting, and worth considering (assuming the review is not horribly biased) for anyone working with loss aversion. It is very likely true that an intuitively appealing conceptualization tends to be overgeneralized and that scientists easily persist even in face of evidence to the contrary.

 


Gal, D., & Rucker, D. (2017). The Loss of Loss Aversion: Will It Loom Larger Than Its Gain? (SSRN Scholarly Paper No. ID 3049660). Rochester, NY: Social Science Research Network. Retrieved on 12 Jul 2018 from https://papers.ssrn.com/abstract=3049660

TIL depression as an unfortunate result of emotional recalibration

(And re: previous post – no, I’m not horribly depressed, nor am I working full-time again. I’m doing things I enjoy in order to get better, and emotion theory happens to be one of them.)

Reading Tooby & Cosmides (2005), Conceptual Foundations of Evolutionary Psychology, as a part of refamiliarizing myself with basics of evopsych. It is a very good description of a lot of basic ideas behind evopsych, having mostly familiar stuff and surprisingly little stuff I disagree with, but the new part I had not run into before was the idea of recalibrational emotion programs.

The core idea is that unlike many other emotions*, emotions such as guilt, grief, shame, gratitude, and depression, have not evolved for producing any immediate behavior change. Instead, drawing from the computational approach to psychology, the idea is that behavior generally is dependent on a lot of (nonconscious) regulatory variables that track the relatively stable circumstances of one’s life. This way the brain does not have to calculate things like the estimate of social support, evaluation of a particular person’s likelihood of reciprocating kindness (or aggression), or present health and energy of own body, on the fly when already in a situation. But these variables need to be updated constantly, and sometimes the act of updating a variable itself should cause changes in other evaluations, in default modes of behavior in related situations, and so on. The authors use guilt as an example (p. 59):

Imagine a mechanism that evolved to allocate food according to Hamilton’s rule, situated, for example, in a hunter-gatherer woman. The mechanism in the woman has been using the best information available to her to weight the relative values of
the meat to herself and her sister, perhaps reassuring her that it is safe to be away from her sister for a short time. The sudden discovery that her sister, since she was last contacted, has been starving and has become sick functions as an information-dense situation allowing the recalibration of the algorithms that weighted the relative values of the meat to self and sister. The sister’s sickness functions as a cue that the previous allocation weighting was in error and that the variables need to be reweighted—including all of the weightings embedded in habitual action sequences. Guilt functions as an emotion mode specialized for recalibration of regulatory variables that control trade-offs in welfare between self and others […]  Previous courses of action are brought to mind (“I could have helped then; why didn’t I think to?”), with the effect of resetting choice points in decision rules.

The authors briefly mention depression as well: “Former actions that seemed pleasurable in the past, but which ultimately turned out to lead to bad outcomes, are reexperienced in imagination with a new affective coloration, so that in the future entirely different weightings are called up during choices.”

I have always considered the functional explanations of depression (or sadness) suspect, because they have typically only briefly mentioned “deattachment” or something similar as the function, and that has sounded… not right. Why would depression have such a horrible feeling if it was simply aimed at deattaching or realigning goals or something? An adaptation that seems to make you passive, drive you to ruin your relationships, and ultimately kill yourself does not seem very adaptive**.

The idea of recalibration makes this so much more understandable! It is not that depression is an adaptation in itself. Instead, (I now hypothesize,) it is the result of the recalibration program accidentally recalibrating all the core motivations at the same time to zero. Normally, the recalibration operates on one core motivation at the time – it gets set to zero, but other motivations are still running, so the behavior is directed more adaptively. But in a case where you only happen to have a couple of core motivations, and they all get set to zero due to some recalibration (that is not maybe based on the most objective of evaluations if the preceding states are already biased), you end up in the state where you have no core motivations left: depression. The program signals that you should not be doing these things, do the other things, but there are no other things left to do.


 

*) The authors strongly subscribe to a discrete emotion view, but their arguments can be read as somewhat inconsistent with the basic emotion theories.

**) Yes, yes, suicide can be adaptive from the gene’s point of view. But depression does not seem a reliable result of situations where suicide would actually be adaptive.

Reference

Tooby, J., & Cosmides, L. (2005). Conceptual Foundations of Evolutionary Psychology. In D. M. Buss (Ed.), The handbook of evolutionary psychology (pp. 5–67). Hoboken, NJ: Wiley.

Learning: measures of political orientation

Writing a meta-analysis on political orientation and MFT, I’ve learned that there are (self-report) measures of political orientation that have completely different approaches to the question, and they seem to be favored by different kinds of users. I’m sure there are interesting philosophical writings on the subject, because topics like “what is political orientation” and “what is ideology” (not to mention “what is measurement”) are probably inexhaustible. This is just my quick observation on the subject.

Four (and a half?) different types of self-report measures:

Self-identification measure uses broad labels (such as “liberal vs. conservative”, “left vs. right”), asking people where they would place themselves (practically always on a bipolar) scale. A pragmatic choice – the focus is somewhere else, but something about the PO needs to be found out, so let’s use the easiest and simplest one. Views PO as a matter of identification and so skips the difficult questions of how to measure PO by settling for the idea that PO is whatever people think it is. However, because the scale predefined but not by the respondents themselves, this may lead to problems when using the measure outside the population from which it emerged – e.g. the use of “lib-cons” and “left-right” are different in the US and Europe. Also psychometrically problematic.

Issue-based measures ask an array of questions about specific political issues. They seem to be favored by researchers in politics – most likely they know about the inconsistencies in people’s identification vs. political behavior (also, their focus is probably more often specifically in political behavior) and the problems with minor groups that do not fit into the bipolar scale, so a single-item scale is viewed as inadequate. Issue-based measures are often combined with self-identification measures to empirically label self-identifications to sets of issue patterns. The approach is that one’s PO is determined by the similarity to other people (a data-driven approach). A general problem with an issue-based measure arises when the issues in the items are not relevant for the respondents (e.g. too old or from a foreign political culture: abortion may be a hot topic in the US and catholic countries, but it is a non-issue in protestant Western/Northern Europe). Another problem emerges when the identifications and response patterns diverge.

Theory-based measure is based on the a priori definitions from particular theoretical approach, and in practice considers the ways people identify themselves as more or less irrelevant. One’s PO is determined by the correspondence to the theoretically important factors that may be issue-based or use items that are about more abstract principles, or both. If you answer in a particular pattern, you can be labeled with a name related to the theoretical thought behind this (such as “liberal vs. authoritarian/statist”), whether you like it or not. I’ve especially seen it in libertarian opinion writings with no empirical research, but it is also used in research. The problem with the former is often that even though a label opposite to their own favored position may seem objective to the writer (like “authoritarian”), people who are not already on the same side surprisingly do not appreciate being  called that, so the writing achieves more to flame than discuss. The tradeoff in the theory-based research approach (in addition to those already related to the issue-based approach) is naturally that often you find what you look for, so the measure can be only as strong as the theory – and I’m not aware of particularly good psychological theories that would take a realistic view of human mind and political behavior into account.

Proxy measures use a measure not intended specifically for political orientation – such as values, particular personality traits like SDO, openness, or moral foundations – but that have earlier established reliable relationships the user can rely on. Psychologists seem to be especially fond of this, probably because they are not interested in political issues per se. Obviously, the problem is that this measure can only capture the facets of PO that happen to correlate with the proxy, so it may miss important information.

In regard to affective psychological view, the different types are not just different ways to answer the same question, but (to some extent) reflect fundamentally different processing. My assumption is that there is a (probably largish) number of nonconscious processes of different levels that produce the range of phenomena that may go under “political orientation”, and that some of the differences in the operating parameters of the low-level processes produce an important portion of the stable differences between political orientations. These trait differences behind political orientation are what I’m mainly interested in. It is not necessarily that the high-level differences are not important, but I assume that their operation in interaction with cultural influences are more complex, and therefore more difficult to study, so it’s better to at least start with the more easier ones.

Self-identification is a result of the conscious processing of the automatic construction of identification categories and those linking oneself to one of them. As such, it is a high-level process, and although that will have some relationship to the trait differences in the lower levels, the heavy processing of the higher levels mostly serve to confound that relationship. So self-identification tells us something about how people self-identify, but less about how this identification fundamentally works – although of course mapping the identification patterns may provide good information for further study. Proxy measures may target the low-level processes better, but they leave the relationship to PO itself unclear, if we assume that PO is something more than just the proxy (and we do).

Issue- and theory-based measures require judgments instead of self-reflection, which is in principle a better way to probe nonconscious processes. The issue-based approach cannot escape the problem that you should have a pretty good theoretical idea of what matters when you choose the items, or else your measure is just a collection of random items that are related to each other on the surface, but that don’t necessarily tell much about the processes underneath. Of course, if the research is on the surface level, that’s completely fine. A theory-based measure with abstract principles comes with the disadvantage that asking people how they would make judgments may give different answers than making them actually make the judgments. So if my assumption about the underlying structure is correct and we want to study the low-level processes with self-report measures, judgment items about issues, guided by a good theory, seems to be the way to go (without spending too much time on analyzing this).

 

TIL several first authors

I was just informed by my colleague that the point of several “first” authors – which is in practice (in our field) used in publication lists to show who has done the most work – is that I could switch the first names in my publication list / resume (perhaps with a note explaining this so that people searching for it don’t get confused) to show that I’m a first author as well. Is this a thing? I’ve never heard of it, but I’ve never understood the point of several first authors either, and this would explain the sense in it. Do people do this?