8  Signaling and social pressures/social identity

8.1 Signaling concern for effectiveness/impact versus other values

Includes, e.g.,

  • social perceptions and ‘signaling empathy vs effectiveness’ ((Caviola, Schubert, and Nemirow 2020))

  • difficulty of coordinated signaling with a hard-to-coordinate measure such as ‘effectiveness’ Burum, Nowak, and Hoffman (2020)

Connection to terms/concepts in other work

Jason Schukraft: Considering all the talk of deontology in this section, I’m surprised you didn’t bring up obligations. I think it’s plausible that many people believe (rightly or wrongly) that they have much stronger moral obligations to their friends/neighbors/community than they do to distant strangers. You can’t discharge your duty to your community by donating to AMF but you can through volunteering at the local soup kitchen.

David Reinstein: This may also drive local giving, however, I don’t see this as a ‘barrier’ or bias. this is simply a particular non-utilitarian ethic.

8.2 Theory and argument

Simler and Hanson (2017) argue that social signaling is a major driver of human behavior, particularly in the charitable domain. Essentially, we may make choices that we would not have otherwise made in order to boost our reputation among our peers, colleagues, etc.. Reputation may be valued for its own sake or instrumentally, as a means to inducing others to act more favorably to us. Game theory (see esp. (Spence 1973)) offers a precise conception of this signaling as a costly way to demonstrate one’s positive “type” in a context with asymmetric information.

Several authors (can be interpreted to) suggest that valuing ‘’effectiveness in generosity’‘, i.e., moral-utilitarianism, is seen as a negative signal by peers and lowers fitness in the cooperation market, at least in comparison to signaling other empathic pro-social responses, perhaps involving a ’virtue ethic’. The exchange below captures the basic argument… (unfold)

From Robin Hanson and Rob Wiblin exchange on 80000 hours podcast; see Chapter 12 of “The Elephant in the Brain” (Hanson)

Hanson: But of course people spend a lot of time directly helping even when they’re relatively well-paid and they could pay other people who earn much lower wages to do a lot more.

Wiblin: This is the example of the high-flying lawyer dishing out soup in a soup kitchen.

Hanson: … the alternative theory that we suggest is that you are trying to show that you feel empathy. That is you want to show there is an emotional capacity in you such that if you see someone around you in need you will feel like you want to do something about that. And existing charities do tend to successful show that. They show somebody who needs help in a direct way that invokes your emotions and you do help to some degree, you do the thing that people would say would help and that shows people around you that you’re not an uncaring person and it might show them, for example, that if they were in need of help later and they were near you you would see them and you would feel about them too. You want to show people that you will be useful ally. If either of you is in trouble the other will come to their aid.

Wiblin: So why is it more important to show the people that you’re the kind of person who if they someone in pain that they’re going to try to help them right then and there than to show that you’re the kind of person who’s smart enough to think about which charities are useful and does their research and actually tries to help people? Because if you don’t care about whether charities are effective or not, my thoughts would just be that you’re not really going to pay attention to whether you’re actually helping your friends or not?

Hanson: Right, but at least if I want your help and I’m your friend it will be my job to put myself in your face and to tell you about my problem. And maybe I figure I coul successfully get myself in front of your face and make you pay attention to my problem and help you understand what I think is effective and then you would just do what I say, and that’s maybe what I’m mostly hoping for. And if you were this person who thinks carefully about how to help the best person in the world who needs help, well I’m plausibly not going to be that best person in the world who needs help so I’m not going to win out in that contest so it’s not actually going to be that useful to know you as the sort of person who will help the person in the world who needs the most help.

(J. A. Everett, Pizarro, and Crockett 2016) argue that deontological ethics signal stable cooperative behavior to others, which enhances fitness in mutualistic partner choice models.1 The basic argument is “An individual who claims [and believes] that stealing is always morally wrong … seems much less likely to steal from me than an individual who believes that the stealing is sometimes morally acceptable depending on the consequences.”

Background and further discussion … in fold

The authors’ motivation is to explore why “intuitive moral judgments often”share characteristics with deontological theories while “consequentialist judgments are often the result of slow, deliberative cognitive processes”. Their key theoretical argument cites “mutualistic partner choice models of the evolution of morality”, which…

posit a cooperation market such that agents who can be relied upon to act in a mutually beneficial way are more likely to be chosen as cooperation partners, thus increasing their own fitness

the typical deontological reason for why specific actions are wrong is that they violate duties to respect persons and honor social obligations-features that are crucial when selecting a social partner. An individual who claims that stealing is always morally wrong and believes themselves morally obligated to act in accordance with this duty seems much less likely to steal from me than an individual who believes that the stealing is sometimes morally acceptable depending on the consequences. Actors who express characteristically deontological judgments may therefore be preferred to those expressing consequentialist judgments because these judgments may be more reliable indicators of stable cooperative behavior.

And recent theoretical work has demonstrated that -“cooperating without looking”—that is, without considering the costs and benefits of cooperation—is a subgame perfect equilibrium (Hoffman, Yoeli, and Nowak 2015). Therefore, expressing characteristically deontological judgments could constitute a behavior that enhances individual fitness in a cooperation market because these judgments are seen as reliable indicators of a specific valued behavior-cooperation.

Hoffman, Yoeli, and Nowak (2015) present an evolutionary game theoretic analysis of an indefinitely repeated game where2

  • Player 1 can publicly ‘look’ to see the cost of cooperation,
  • Player 1 next chooses to cooperate or defect, and then
  • Player 2 chooses whether to continue repeating the above game, or end the relationship.

They provide conditions under which ‘cooperating without looking’ (CWOL) is part of a subgame-perfect Nash equilibrium, and an evolutionarily stable equilibrium can involve a substantial rate of CWOL play.

However, the analogy to effective versus ineffective giving is not clear. Perhaps a connection could be made if considering the charity effectiveness tended to provide a motivation to give less, but this is not obviously the case. In general, an ‘excuse not to do something’ is not the same as a ‘choice to be effective’.

8.3 Evidence for this barrier

We consider specific empirical claims that could be interpreted as evidence that this is an important barrier.

Evidence that “consequentialist choices lead to negative signals and less-favorable treatment relative to deontological/emotionally intuitive choices”

Although altruism is generally rewarded, several studies suggest that effective altruism is not

Citing: Burum et al, 2020; Montealegre et al, 2020; Yudkin et al, 2019

(J. A. Everett, Pizarro, and Crockett 2016)

J. A. Everett, Pizarro, and Crockett (2016) ran a series of experiments on the Mturk platform, involving hypothetical dilemmas paired with low-stakes (or no) Trust games.

In each, participants were asked to make and justify their judgement in a moral dilemma such as the famous ‘trolley dillemma’. In each case, this was a (hypothetical) choice between inaction and taking an action that sacrifices a small number of lives to save a larger number of lives.3

We then had our participants rate the morality and trustworthiness of each agent on a scale (Study 1a), play a hypothetical trust game (TG) with the agents (Study 1b), and, finally, play a TG involving real monetary stakes with the agents (Study

Across several studies deontologist agents were preferred in partner choices by participants endorsing the same values; even moral-utilitarians seem to favor peers who express emotional empathy and deontological ethics (ibid., p. 45).

However, this difference was not present in the track-switching dilemma, the only dilemma in which a majority favored the consequentialist choice.

Incorporate some of the other papers mentioned in the Gdoc here

Montealegre et al. (2020) Does Maximizing Good Make People Look Bad? Manuscript

Type of evidence: Online experiment (M-turk) with hypothetical choices (attitudes), six pre-registered studies using two different scenarios (N = 1,961)

Relation to charitable giving: directly, since …

Scenarios (including within and between subjects approaches):

  1. participants evaluate a response to survey question about how to select a charity (“If you were to donate, how would you select which charity to donate to?”).

Deliberative: I would use evidence to calculate which charity spends its donations most cost effectively, and donate to them.

Empathic: I would try to put myself in the shoes of people who are going through difficult situations, and donate to a charity that helped them.

  1. John was approached by a charity fundraiser and was asked whether he would be interested in donating to help Rokia, a 7-year-old girl from Mali, Africa who was desperately poor and faced the threat of severe hunger or even starvation. The charity fundraiser showed John a picture of Rokia (also presented to participants) and then asked John whether he would be interested in supporting her.

Across conditions, participants were presented with the general scenario and only the potential donor’s actions differed depending on the condition:

Deliberation: “John thought that donating to help Rokia might not be the most cost-effective way to use his money and that maybe he should donate to a charity doing something more cost-effective instead. He asked the charity fundraiser about the relevant statistics of the program and since the data suggested this charity was the most cost-effective John donated to the charity.”

Empathy: “John was deeply moved by Rokia’s situation and about how terrible her situation must be for her. After hearing about her tragic story and imagining how his donation could help her John donated to the charity.”

Robustness checks for: gender, stake size

Background mechanism: donors’ failure to prioritize cost-effectiveness can be explained by signaling concerns, since people who favor deontological over consequentialist decisions are preferred as social partners (J. A. C. Everett, Pizarro, and Crockett 2016)

Key findings: Across six studies, donors who deliberated were perceived as having worse moral character, were rated as less desirable as social partners, and were judged to be less guided by moral motives. On the other hand, those who deliberated were also seen as more reasonable and competent, and were judged to be more guided by pragmatic motives. Thus, there may be reputational benefits associated with deliberating. However, since deliberators are less preferred as social partners the authors believe the overall effect one reputation is negative. They do not find any differences in trustworthiness. Exhibiting empathy before deliberating reduced most negative reputational effects. People how do not give at all are seen as the worst. Surprisingly, participants judge the empathy donors as more wasteful.

A critique (Jason Schukraft):

This set-up seems misleading to me because in the first case it appears to be left uncertain whether John would donate to a different (more effective) charity if the data didn’t pan out.

It also might imply a false dichotomy: in the first case, readers might imagine that John doesn’t experience an empathetic response? In my view, EA is successful because it combines empathy with deliberation. I would be very suspicious of an EA who wasn’t empathetic.

Burum et al, ’20

(Burum, Nowak, and Hoffman 2020)

Involving variation in donation multipliers and social signaling/rewards4

  • Yudkin et al, 2019, “Actions speak louder than outcomes in judgments of prosocial behavior”

  • Johnson https://psyarxiv.com/r85jv/ … Mturk experiments, people asked about praiseworthiness of hypothetical donors, between subject variation in amount donated and amount of impact achieved (1 of 6 studies)

  • Kawumara et al, 2020, perhaps in contrast to the above, seems to suggest that ‘effectiveness’ is rewarded with reputation. he people in the vignettes did seem to be rewarded (assessed as having both higher warmth and competence) by the participants for achieving a greater benefit, holding cost constant


Even if we accept the above evidence (that those who make consequentialist active choices in “sacrificial dilemmas” are seen as less trustworthy and less moral) this may not generalize to effective giving. Researching and selecting a more effective charity is closest to the ‘track switching’ scenario in these experiments, in which no substantial difference was observed, and even here it is a stretch. Choosing a more effective charity instead of a more local one (e.g., river blindess prevention in Africa versus local guide Dogs) would be hard to cast seen as taking an active step to harm someone.5

Indirect evidence

(Kahane et al. (2018); Jordan et al. (2016); Hoffman, Yoeli, and Nowak (2015))

  1. Jason Schukraft: I think the consequentialism vs deontology framing is a bit misleading. I think there’s an implicit virtue ethics at work here. Volunteering at the soup kitchen demonstrates the possession of certain virtues, and these virtues can’t be adequately signalled by donating to effective charities. (Donating to effective charities evidences other virtues, of course.)↩︎

  2. Here the analogy to effective versus ineffective giving is not clear…↩︎

  3. Would you “push a man off a footbridge to stop an oncoming train from hitting five”?↩︎

  4. Writeup needed here↩︎

  5. True, a local blind person may fail to get an additional (fraction of a) dog as a result of this choice. However, there is little sense in which “you funding his dog” would be seen as the status quo absent your intervention.↩︎