EMPATHY AND TECHNOLOGY

DEFINITIONS, EXAMPLES, AND FUTURE DIRECTIONS

Presentation by Christine Rosakranse, Stanford University Department of Communication

Prepared for Facebook HQ, December 8, 2016

THE GOAL OF EMPATHY RESEARCH IS TO…

FIND A WAY TO MAKE PEOPLE MORE COMPASSIONATE OR TO FOSTER EMPATHY AND COMPASSION THROUGH AN EXISTING TECHNOLOGY

EMPATHYWHAT IS IT?

TEXT

THEORY OF MIND

▸ Feeling another’s emotions

▸ Cognitive perspective-taking

▸ Mirror Neurons and ToM (Trump Effect)

▸ Autistic example

EMPATHY: ITS ULTIMATE AND PROXIMATE BASES, (PRESTON, DE WAAL)

▸ Ultimate bases: “Empathy increases with:

Familiarity

(subject's previous experience with object)

SIMILARITY

▸ Similarity can be manipulated (red team vs. blue team)

(perceived overlap between subject and object e.g. species, personality, age, gender),

LEARNING

▸ Learning

(explicit or implicit teaching)

PAST EXPERIENCE

▸ Past experience

▸ (with situation of distress)

SALIENCE

▸ (strength of perceptual signal e.g. louder, closer, more realistic etc.)

MOTIVATIONAL THEORIES OF EMPATHY

BATSON’S EMPATHY-JOY-ALTRUISM HYPOTHESES

EMPATHIC-JOY HYPOTHESIS

▸ “We reasoned that if empathically aroused individuals are egoistically motivated to gain empathic joy, then their desire to hear from the needy person again should be a direct function of the likelihood of obtaining empathic joy…”

▸ Empathic-altruism: “altruistically motivated individuals should want to hear how the needy person is doing even when the chances of improvement are not great.”

EMPATHIC-JOY HYPOTHESIS VS. EMPATHY-ALTRUISM

BATSON ET AL.’S ABSTRACT

▸ Three experiments tested whether empathy evokes egoistic motivation to share vicariously in the victim's joy at improvement (the empathic-joy hypothesis) instead of altruistic motivation to increase the victim's welfare (the empathy-altruism hypothesis).

▸ In Experiment 1, Ss induced to feel either low or high empathy for a young woman in need were given a chance to help her. Some believed that if they helped they would receive feedback about her improvement; others did not.

▸ In Experiments 2 and 3, Ss induced to feel either low or high empathy were given a choice of getting update information about a needy person's condition. Before choosing, they were told the likelihood of the person's condition having improved—and of their experiencing empathic joy—was 20%, was 50%, or was 80%.

▸ Results of none of the experiments patterned as predicted by the empathic-joy hypothesis; instead, results of each were consistent with the empathy-altruism hypothesis.

Perspective-taking as a method for increasing empathy

PERSPECTIVE-TAKING

▸ “To manipulate empathy, some subjects were asked to adopt an objective perspective while watching (low-empathy condition), and others were asked to imagine how the young woman felt (high-empathy condition).”

IMAGE FROM PRESTON, DE WAAL HTTP://COGPRINTS.ORG/1042/1/PRESTON_DE_WAAL.HTML

FOR FIRST EXPERIMENT

DEPENDENT MEASURES▸ Dependent measure: Volunteering to help Katie418 BATSON, BATSON, SLINGSBY, HARRELL, PEEKNA, TODD

dence from these two checks, we concluded that the empathymanipulation was successful.

Relationship of self-reports of empathy, distress, and sadness.In keeping with the measurement issues raised earlier and theprevious observations of the relative indistinguishability of self-reports of empathy and distress in response to the need situa-tion used in this study (Batson et al., 1989; Batson et al, 1988), aprincipal-components analysis of responses to the six empathyadjectives, eight distress adjectives {alarmed, grieved, troubled,distressed, upset, disturbed, worried, and perturbed), and foursadness adjectives {low-spirited, feeling low, heavyhearted, andsad) revealed that with the exception of perturbed, all of theseadjectives loaded on a single component. Omitting perturbedfrom the list produced a one-component solution (eigenvalue =11.13). All 17 adjectives loaded .60 or higher on this component.This one-component solution suggested that in response to thepresent need situation, the so-called empathy, distress, and sad-ness adjectives actually measured a single underlying dimen-sion.

One interpretation of this dimension was, as Batson et al.(1989) and Batson et al. (1988) suggested, that the distress andsadness adjectives were at least as good an index of empathicfeelings for Katie as were the empathy adjectives. This interpre-tation seemed even more plausible in light of the highly signifi-cant main effect of the empathy manipulation on an index cre-ated by averaging responses to all 17 adjectives (Ms = 3.14 and4.92 for the low- and high-empathy conditions, respectively),F{\, 66) = 51.41, p< .0001 (Fs < 1.0 for the feedback manipula-tion and the interaction). Other interpretations were that thesingle dimension reflected general arousal or either distress orsadness. In any case, the lack of clear differentiation betweenempathy and distress responses called into question the appro-priateness of creating an empathy-minus-distress index in re-sponse to this need situation.

Effect of anticipated feedback on the empathy-helping rela-tionship. Helping responses were coded in two ways. First, theproportion of subjects who volunteered any amount of timeserved as a dichotomous measure of helping (0 = no help, 1 =help); second, scores on the 4-point scale of number of hoursvolunteered {0 = 0hr,l = 3-5 hr, 2 = 6-8 hr, and 3 = 9-10 hr)served as a continuous measure. Scores on the continuous mea-sure were badly skewed (29 subjects did not help, 33 subjectsvolunteered 3-5 hr, 5 subjects volunteered 6-8 hr, and 5 subjectsvolunteered 9-10 hr) and indicated an underlying dichotomy ofresponses. Therefore, we adopted the dichotomous measure asour major index of helping. As a check, we also analyzed thescaled measure.

Table 1 contains the proportion of subjects volunteering tohelp Katie in each cell of the 2 (low empathy, high empathy) X 3(no information, no feedback, feedback) design. Means on thescaled measure are also reported.

An analysis of variance on the proportion of subjects volun-teering to help Katie in each cell (normal approximation usingthe arcsine transformation; see Langer & Abelson, 1972; Winer,1971) revealed a reliable empathy main effect, x2(l, N=12) =5.04, p < .025. This main effect was, however, qualified by amarginally significant interaction, x2(2, N= 72) = 4.87, p <. 10.Inspection of cell means and tests for simple main effects re-vealed evidence of the empathy-helping relationship predicted

Table 1Proportion of Subjects Agreeing to Help Katiein Each Cell of Experiment 1

Empathycondition

LowProportionM

HighProportionM

No informationabout feedback

.420.67

.751.00

Information about feedback

No feedback

.330.33

.831.17

Feedback

.670.92

.580.75

Note, n = 12 per cell. The means are those for the scaled measure ofhelping, ranging from no help (0), 3-5 hr (1), 6-8 hr (2), to 9-10 hr (3).

by the empathy-altruism hypothesis in both the no-informa-tion (replication) condition (z = 1.69, p < .05, one-tailed) and theno-feedback condition (z = 2.62, p < .005, one-tailed). Thesignificant difference in the no-feedback condition was con-trary to the prediction of the empathic-joy hypothesis.

Unexpected by both the empathic-joy hypothesis and the em-pathy-altruism hypothesis, there was no evidence of an em-pathy-helping relationship in the feedback condition. We con-ducted further analyses to determine why. Although the rate ofhelping in the high-empathy/feedback cell was somewhat lowerthan the rate in the other two high-empathy cells, neither ofthese differences approached statistical significance (both zs <1.40). Instead, the lack of an empathy-helping relationship inthe feedback condition seemed to be due primarily to the rela-tively high helping among the low-empathy subjects led to antic-ipate feedback; there was a marginally significant increase inthe proportion of low-empathy subjects helping in the feedbackcondition (.67) compared to the no-feedback condition (.33; z =1.66, p < .10, two-tailed). Rather than the helping of high-em-pathy subjects dropping to the level of low-empathy subjects inthe no-feedback condition, as had been predicted by the em-pathic-joy hypothesis, the helping of low-empathy subjects roseto the level of high-empathy subjects in the feedback condition.Low-empathy subjects, not high-empathy subjects, were theones whose helping increased with anticipation of feedback.

In retrospect, this unexpected increase seemed entirely rea-sonable. Empathic emotion is, of course, not the only source ofmotivation to help, and it seemed reasonable that low-empathysubjects, less focused on Katie's welfare than high-empathy sub-jects, would be more focused on the potential social and self-re-wards for helping, including the vicarious pleasure of knowingKatie was better. Perhaps they, rather than high-empathy sub-jects, had, in Smith et al.'s (1989) words, "enhanced sensitivity tovicarious joy" (p. 642), and so helped "in order to be happy"(p. 641).

Finally, note that contrary to the assumption of Smith et al.(1989), subjects' responses in the no-information condition par-alleled those of subjects explicitly told that they would not re-ceive feedback more than those of subjects told that they wouldreceive feedback. The scaled measure of helping patterned ex-actly as did the dichotomous measure, although effects weresomewhat weaker.

An analysis of variance on the proportion of subjects volunteering to help Katie in each cell revealed a reliable empathy main effect, x2(l, N=12) = 5.04, p < .025. This main effect was, however, qualified by a marginally significant interaction, x2(2, N= 72) = 4.87, p <. 10.

WILL EMPATHY INTERVENTIONS WORK?

ONLINE?

ONLINE RAMIFICATIONS FOR INTERPERSONAL EMPATHY

▸ Computer-mediated communication eliminates some real-world context clues for sharing emotion

▸ However, it can also be hyperpersonal (Walther)

▸ Hian, Chuan, Trevor, and Detenber's 2006 study found that “relational intimacy” increased at a faster rate in CMC than in Face-to-Face interactions

MEDIA AGAINST BIAS

➤ Mixed results

➤ VR example:

➤ Putting Yourself in the Skin of a Black Avatar Reduces Implicit Racial Bias (Tabitha C. Peck, Sofia Seinfeld, Salvatore M Aglioti, Mel Slater)

➤ The influence of racial embodiment on racial bias in immersive virtual environments (Victoria Groom, Jeremy N. Bailenson, and Clifford Nass) - increases bias in IAT

DIGITAL MEDIA EFFECTS

➤ Heavy social media use can be both an indication of narcissism and a means of supporting and perpetuating narcissistic behavior (Buffardi & Campbell, 2008; Leung, 2013)

➤ Social networks tend to be “homogenous with regard to many sociodemographic, behavioral, and interpersonal characteristics” (McPherson et al. 2001)

➤ Face-to-face communication is not the same as computer-mediated communication (Walther et al., 1994; Riva, 2002)

MY EXPERIMENT

COMBINES THREE EXPERIMENTAL APPROACHES

➤ Oswald’s manipulation: designed a study using videotape to demonstrate the relationship between affective empathy, cognitive empathy, empathic concern and altruistic helping

➤ Decety and Yoder’s questionnaire design: using affective, cognitive, and social justice sensitivity as measures

➤ Imagined intergroup contact paradigm: recording written responses as a method for measuring perspective-taking

INDEPENDENT VARIABLES

➤ 2 x 3 design: Empathy condition by cued condition

➤ Empathy condition: Affective, cognitive, or distraction

➤ Cued condition: Cued or not cued for comment

Affective Cued

Cognitive Cued

Distraction Cued

Affective Not Cued

Cognitive Not Cued

Distraction Not Cued

DEPENDENT VARIABLES

➤ Bias Measures

➤ Homophobia

➤ Resentment and Over Racism Against African-Americans

➤ Illegal Alien Scale

➤ Difference in Justice Sensitivity Scales

➤ Helping Behaviors

➤ Volunteering

➤ Money Donation

ADDITIONAL COVARIATES

Gender, Narcissism and Interpersonal Reactivity

HYPOTHESES

➤ H1a-g: There will be a main effect of empathy conditions on bias against marginalized groups.

➤ H2a-g: The interaction between empathy condition and cued condition will result in significant effects on the bias measures.

➤ H3a-d: There will be a main effect of empathy condition on justice sensitivity.

➤ H4a-d: The interaction between empathy condition and cued condition will result in significant effects on justice sensitivity.

➤ H5a-d: The interaction between empathy condition and cued condition will result in significant effects on helping behavior.

RESULTS(but first a little thought experiment)

This is George…

RESULTS: H1 - MAIN EFFECT OF EMPATHY CONDITION ON ILLEGAL ALIEN SCALE

F (2, 135) = 3.31* p = .039

H2 - WHAT IS AN INTERACTION EFFECT?

An interaction effect occurs

when the the effect of one

independent variable depends

on the level of the other

independent variable.

H2A-D: EFFECTS ON HOMOPHOBIA

F (2, 135) = 3.64* p = .029

Factor 1,

Behavior/Negative Affect

H2A-D: EFFECTS ON HOMOPHOBIA

F (2, 135) = 3.34* p = .038

Total:

Homophobia

JUSTICE SENSITIVITY

with 4 subscales

Findings for Justice Sensitivity, 4 subscales

JUSTICE SENSITIVITY: FACTOR 1, SELF-ORIENTED (VICTIM)

F(2, 136) = 4.63 p = .01

JUSTICE SENSITIVITY: FACTOR 3, OTHER-ORIENTED

F(2, 135) = 3.63* p = .03

HELPING BEHAVIORSVolunteering and money donation

VOLUNTEERING: CIVIL RIGHTS AWARENESS BOOTH

F(2, 135) = 3.09* p = .049

SIGNIFICANCE OF FINDINGSWhen empathy succeeds and why it fails

EMPATHY VERSUS “RATIONAL COMPASSION”

➤ Implication of the medium

➤ Not the same as face-to-face

➤ Increasing interactivity has different effects across conditions

➤ The popular YouTube format and digital storytelling may have adverse effects on bias against marginalized

➤ Compounded by empathy prompts

➤ Affected by forcing response

RELATED TO THE POPULATION

➤ Least bias against illegal aliens in the distraction condition

➤ Least homophobia in the distraction cued condition

➤ Empathy intervention activated self-centeredness

➤ Justice sensitivity for the self increased in the affective condition

➤ Distraction condition had more other-oriented thinking

➤ Affective, cued had most volunteering for civil rights booth

NEXT IN RESEARCH

➤Empathy

Core is a person’s initial attachment, self-esteem, level of narcissism, natural ability.

Learning can expand an initial self-interest through active listening and other exercises.

DIRECTIONS

➤ Looking at empathy as one tool for bias reduction

➤ Reduce narcissism

➤ Help practice non-judgement

➤ Strengthen core: self-esteem, ability to perform emotional labor

➤ Researching the role of collaboration in reducing bias

➤ As a means of expanding self-interest

➤ Creating a greater sense of common humanity

ADDITIONAL RESOURCES

▸ Ashoka

▸ Practical ideas for solving social problems

▸ Non-Violent Communication (NVC)

▸ Active listening

▸ Mindfulness

▸ Humanistic Values

▸ d.school at Stanford for empathy in the design process

THANK YOU.