NOTE-1274: Empirical Approaches to Altruism

Contribution

A reference survey of the empirical case for psychological altruism. It fixes the target with the standard account: altruism is an ultimate desire for another’s well-being, and egoism says all ultimate desires are self-interested. It sets aside the biological and social-science senses of the word. It then takes Batson’s empathy–altruism programme as the best empirical work, and reports what it has and has not ruled out. For the record, its value is §5.3. That section states the self-other merging challenge, quotes the measures both 1997 papers used, and argues that none of them measures what the challenge needs.

Key insight

The relevant identity is numerical, not qualitative. Saving for one’s own old age is egoistic because old Sophia is young Sophia, not because she resembles her (§5.3). So a merging test matters for egoism only if it shows that A takes herself to be numerically identical with B. Perceived similarity and overlap in trait ratings measure likeness. “It is far from clear what the other two tests [IOS and use of ‘we’] are measuring” (§5.3). The authors conclude that the merging literature gives no evidence that people lose the distinction between themselves and others. Its remaining challenge is causal: whether empathy or “merging” drives the helping.

Assumptions

• The standard account (§2). Ultimate desires are those not produced by practical reasoning from other desires. Egoism and altruism are claims about them. Some desires are neither: revenge, duty, preserving art.
• Numerical identity is the self’s extent that matters (§5.3). The authors give no argument that an extended self-concept short of identity could make a desire self-interested. They do not discuss it.
• Batson’s conception of empathy: “an other-oriented emotional reaction to seeing someone suffer” (Batson 1991: 58, quoted in §5), distinct from self-oriented personal distress.

Key results

• Egoism–altruism and evolution (§3). Biological altruism lowers the actor’s fitness. It neither entails nor is entailed by psychological altruism (Sober and Wilson 1998). Piccinini and Schulz (2019) add a second axis, how a desire was produced. An ultimate desire for another’s good that was acquired by reward learning is altruistic only “non-classically”.
• Social punishment (§5.1; Fultz et al. 1986, Study 2). Lonely “Janet”; empathy by perspective-taking instructions; anonymity versus scrutiny. Mean time offered, as the table prints it: high scrutiny 0.67 (low empathy) and 1.71 (high empathy); low scrutiny 1.29 and 2.44. High empathy raised helping under both, against the social-evaluation hypothesis.
• Aversive-arousal reduction (§5.2; Batson, Duncan et al. 1981, Experiment 1). Taking “Elaine’s” shocks. Proportion helping: easy escape 0.18 (low empathy) and 0.91 (high empathy); difficult escape 0.64 and 0.82. Empathy was manipulated by similarity of values. The same pattern held across six experiments, and broke only when the shocks were “clearly painful”. Stocks et al. (2009) manipulated the ease of psychological escape and found empathy still predicted helping.
• Self-other merging (§5.3).
  • The challenge, after Aron and Lerner: in closeness, perspective-taking or empathy, the self–other boundary may vanish. Batson concedes that if it does, “so does the distinction between altruism and egoism” (2011: 148, quoted).
  • Cialdini et al.’s oneness index (1997: 484, quoted): the IOS scale of seven increasingly overlapping circle pairs, plus a 7-point item on how far one would use “we” of the relationship.
  • Batson, Sager et al.’s three measures: the IOS scale; perceived similarity, “How similar to you do you think the person …is?” (1 = somewhat, 9 = extremely; 1997: 500, quoted); and the mean absolute difference between self- and other-ratings on personal attributes (1997: 498, quoted).
  • The verdict: none of the four tests shows perceived numerical identity. Someone who believed he existed in two bodies would be delusional (May 2011b: 32, as cited).
  • The causal dispute: Cialdini et al. report three studies in which empathy plays no causal role and “merging”, “a label for whatever the IOS scale and the use-of-we test measure”, drives helping. Batson et al. “pointed out a number of methodological problems”, and “when they conducted a pair of experiments that avoided these methodological problems, the role of empathy in producing helping behavior was clearly evident”. The entry names neither the problems nor the numbers.
  • Robustness: the studies are small. McAuliffe et al. (2018), N = 680 online, preregistered, were “unambiguously supportive” (2018: 504, as quoted) of the empathy–helping link. Note 16 adds that their evidence suggests perspective-taking instructions may not raise empathy; “remain objective” instructions lower it.
• What is still open (§§5.4, 6 [printed as 7], 8).
  • A divine punishment hypothesis: secrecy does not hide a choice from an omniscient God, so the Janet studies cannot exclude it.
  • Principlism, an ultimate desire to uphold a moral principle, which Batson grants is untested.
  • Ultimate desires to comply with any culturally acquired norm.
  • Habitual or Pavlovian helping, not driven by ultimate desires at all (Gęsiarz and Crockett 2015).
• Development (§7). Eighteen-month-olds help without reward, sometimes at a cost, and rewards can crowd it out (Warneken and Tomasello). These studies control egoistic confounds far less than Batson’s.

Claims

id	claim	strength	support
C1	Biological altruism and psychological altruism neither entail nor exclude each other	strong: conceptual, after Sober and Wilson	§3
C2	Batson’s experiments make the social-punishment and aversive-arousal versions of egoism much less plausible than the empathy–altruism hypothesis, in situations like those studied	moderate to strong: many experiments, reported in summary; small samples, one large preregistered replication of the empathy–helping link	§§5.1–5.3
C3	Only perceived numerical identity with the helped person would make empathic helping egoistic	moderate: argued from one example; extended self-concepts are not considered	§5.3
C4	None of the merging measures in the 1997 exchange tests for numerical identity	strong given C3: two measure likeness, two are of unclear content	§5.3
C5	Batson et al. (1997) answered Cialdini et al.’s causal claim by fixing methodological problems, and found empathy’s role evident	weak as reported here: stated in one sentence, with no design, measures or results	§5.3
C6	Altruism is not yet established, because principlism, norm compliance, divine punishment and habitual or Pavlovian helping are not excluded	moderate: conceptual; the divine-punishment alternative is untested in either direction	§§5.4, 6, 8

Concepts

• standard account: altruism is an ultimate desire for another’s well-being; egoism, that every ultimate desire is self-interested.
• ultimate desire: a desire not produced or sustained by practical reasoning from other desires.
• numerical versus qualitative identity: being the same thing, versus being alike.
• self-other merging: the hypothesis that in closeness or empathy the conceptual boundary between self and other disappears.
• principlism: motivation with the ultimate goal of upholding a moral principle (Batson 2011: 220, quoted).
• classical versus non-classical altruism (Piccinini and Schulz): an ultimate desire for another’s good from an altruistic or neutral mechanism, versus from an evolutionarily selfish one such as reward learning.

Connections

• Batson et al. 1997 (LIT-1700) and Cialdini et al. 1997 (LIT-1728), both Unavailable. §5.3 quotes each paper’s measures with page numbers. That is the record’s only first-hand text from either.
• Aron et al. 1991 (LIT-1706). §5.3 quotes “Much of our cognition about the other in a close relationship is cognition in which the other is treated as self or confused with self—the underlying reason being a self-other merging”, attributed to “Aron et al. 1980: 242”. The bibliography has no 1980 Aron entry, and the 1991 paper runs pp. 241–253. So the quotation is probably from LIT-1706, p. 242. This is an inference.
• Sober and Wilson (LIT-1683) supply the standard account and the biological/psychological distinction (§§2–3).
• Markus and Kitayama (LIT-1703) open §5.3 as evidence that self-concepts vary across cultures.
• Stich, Doris and Roedder 2010, the handbook chapter this entry borrows from, is cited for the detailed methodological critique of Batson. A search of its text found no discussion of merging or oneness.
• Miyazono and Inarimori 2021 (read the same day) answer May’s objection, which §5.3 relies on, with a group-identification reading of merging.

Bearing on the record

• Batson’s objection (CLAIM-494). The entry is the record’s first independent report of how the 1997 exchange came out. It sides with Batson. But it gives the result in one sentence, with no design or numbers, so it does not let the record check the claim. One discrepancy needs checking: LIT-1700’s summary and CLAIM-494 say “four experiments”, and the entry says Batson et al. “conducted a pair of experiments”. The claim’s wording should not repeat “four” until the paper is read.
• The owner’s thesis (CLAIM-491). §5.3 bears on it at two points. These are inferences; the entry does not discuss the thesis.
  • It rules out the strongest version of the thesis. If “a self that includes the other” means that one takes oneself to be the other, nobody normal does, and the measures do not show it.
  • It leaves the thesis’s own version unexamined. The thesis says the self-identification extends, not that the helper is numerically one with the helped. The entry’s test of relevance, numerical identity, assumes that only one’s own numerically identical person can be the object of a self-interested desire. The thesis denies exactly that, so the entry does not refute it. It does show that the thesis must say why an extended self-concept counts as “self” for the purpose of interest, since Sophia’s case gets its force from identity.
• The triviality objection (CLAIM-492). The entry supplies one criterion of the self’s extent that plainly comes apart from caring, numerical identity. It is too strong for the thesis to use. It calls the IOS scale and the “we” item, the criteria the thesis and its objection both name, unclear in content. So it sharpens the objection: the record holds no validated criterion of extension between identity and caring.
• What happened when Batson controlled for merging. From this entry: three merging measures, including Cialdini’s IOS scale; correction of unnamed methodological problems; empathy’s role “clearly evident”. Nothing more. The question stays open until LIT-1700 is read or a fuller secondary account is found. Miyazono and Inarimori point to Batson 2018, chapter 10, for the methodological problems.
• A dated note to CLAIM-494 is proposed:

  Note of 2026-10-29: the SEP’s report. Doris, Stich, Walmsley and Schulz (NOTE-1274) report that Batson et al. found methodological problems in Cialdini et al. and, in “a pair of experiments” that avoided them, found empathy’s role in helping “clearly evident”. They give no numbers, and the paper stays unread. They also argue that none of the four merging measures in the exchange tests for perceived numerical identity, so neither side has measured the self’s extent in a way that bears on egoism.

• No THEORY filed. No ML instruction.

Limitations

• A survey. It reports results second-hand and selectively. For the 1997 exchange it gives measures, not findings.
• No pagination. Citations are to sections.
• The numerical-identity argument rests on one example and does not consider graded or extended self-concepts.
• The main page and the notes were saved separately. The scratchpad copy lacked the notes.
• Typos and slips.
  • “Cialdini et al. (1977)” for 1997 (§5.3).
  • “Aron et al. 1980: 242” for what is probably 1991 (§5.3).
  • The section after §5.4 is headed “7. Beyond Egoism vs. Altruism” though the contents list it as 6, so there are two §7s.
  • §3 refers twice to “table 1” for Piccinini and Schulz’s framework. No such table is in the entry; its Table 1 is the Fultz predictions.
  • “the well-being of the that person” (§2); “reassuring in this respect to” and “that that emerged” (§5.3).
  • In-text “Sripada & Stich 2006” against a bibliography entry of 2007. Note 21 cites “Stich & Donaldson 2019” against a bibliography entry of 2018. Note 3 cites “Mill 1861 [2001” against 1861 [2011].
  • Note 21 discusses “the Demjanuk case”, which the main text no longer mentions. It is a remnant of an earlier version.
  • In Table 3 the low-scrutiny cells exceed the high-scrutiny ones in both empathy conditions. Neither hypothesis predicts that, so the row labels may be swapped. The table is reported here as printed; the empathy effect holds either way.
  • “PhilPapers.ord” under Other Internet Resources.

Open questions

• What were the methodological problems Batson et al. found, and how did their two experiments control merging: statistically, by manipulation, or both? Batson 2018, chapter 10, may say.
• Is there a criterion of self-extension between numerical identity and caring that bears on whether a desire is self-interested? Neither this entry nor May (as reported) considers one.