Social Robots, Cross-cultural Differences

Ana R. Delgado and Margarita G. Marquez

Psicología Básica, Psicobiología y Metodología, Universidad de Salamanca, Salamanca, Spain

Keywords: Anger, Contempt, Cross-cultural Differences, Disgust, Moral Emotions, Social Robots.

Abstract: The study of emotion abilities is of interest to Artificial Intelligence because identifying and responding

appropriately to the affective states of humans is thought to make users more prone to interact with robots.

However, cross-cultural differences in social communication are common. The CAD (Contempt, Anger,

Disgust) hypothesis proposes that these three emotions are elicited by different violations of moral codes.

Our exploratory study of texts from a corpus of Spanish contextualized words shows that both the emotion

receiver and its perceived cause are different for these emotions: disgust takes as its object mostly

something concrete, anger is preferentially felt towards another person, and contempt towards an abstract

object. In Spain, disgust was associated with prejudice, and anger with altruistic motives while contempt

remained the most elusive of the triad. In Latin America, both disgust and contempt were associated with

prejudice, while the altruistic function of anger failed to reach significance. Differences concerning the

moral functions of anger and contempt corroborate that the cultural context in which emotions are expressed

can change their moral meaning. The procedure is an ecologically valid one that can be of help for

designing more realistic social robots.

1 INTRODUCTION

In Robotics, emotional skills are assumed as

convenient: recognizing and responding to the

affective states of humans is thought to make users

more prone to interact with robots (Cañamero,

2005). The affective revolution in Psychology is

producing enough knowledge to help engineers to

decide the “best” emotion for different human-robot

interaction situations by taking into account the

evolutionary roots of emotions as well as the

functionality of each one of them in the current

interaction contexts (Delgado, 2009a).

Affects are playing an increasing role in the

newest theories of morality (Bloom, 2010; Haidt,

2008). This approach is partly related to the spread

of evolutionary psychology in which the origins of

morality are related to reciprocal altruism (Cosmides

and Tooby, 2005). The CAD (Contempt, Anger,

Disgust) hypothesis, proposes that these emotions

have distinct moral functions, i.e., would be elicited

by violations of different codes (Rozin et al., 1999).

Rozin et al., (2009) have recently analyzed how core

disgust, which originates in the mammalian bitter

taste rejection system, became true disgust when the

eliciting category was enlarged to include a disgust

evaluation system that responded to more

cognitively elaborated appraisals and, later on, to

social stimuli. Moral offences would be a further

extension of the disgust evaluation system.

However, empirical evidence indicates that disgust

may also be elicited by violations of the autonomy

code, which according to the CAD hypothesis would

be associated with anger (Rozin et al., 1999). The

fact that some moral transgressions trigger the same

facial motor activity that is evoked by distasteful and

basic disgust items (Chapman et al., 2009) is a weak

corroboration of the “from oral to moral” disgust

exaptation hypothesis because it is compatible with

alternative views (Rozin et al., 2009).

As to appraisals or attributions, both reciprocal

and altruistic scenarios are currently considered in

the scope of moral psychology (Cosmides and

Tooby, 2005); (Haidt, 2008). In this sense, results

linking disgust to prejudice indicate that the usual

role of disgust is not the moral one (Taylor, 2007)

and, apart from some isolated observations,

ecological evidence for the ethical role of contempt

is lacking.

From a methodological point of view, moral

psychology has been criticized for focusing nearly

exclusively on studies in which volunteers solve

artificial moral dilemmas imagined by other people

R. Delgado A. and G. Marquez M..

Social Robots, Cross-cultural Differences.

DOI: 10.5220/0004383300190022

In Proceedings of the 15th International Conference on Enterprise Information Systems (ICEIS-2013), pages 19-22

ISBN: 978-989-8565-61-7

 2013 SCITEPRESS (Science and Technology Publications, Lda.)

(Bloom, 2010). The lack of stimulus

representativeness makes generalization to everyday

life problematic (Baumard and Sperber, 2010);

(Delgado, 2009b).

Analyzing representative text corpora allows the

researcher to avoid the previous critiques because

words can be observed unobtrusively (Bauer and

Aarts, 2000); (Webb et al., 1981). A clear limitation

of qualitative text analysis is subjectivity, which can

be somewhat controlled by indirect means, but never

eliminated. In contrast, a clear advantage of this

procedure is that if the text corpus is not in English

then the limitation that researchers might be

imposing concepts from English vocabularies on

participants is avoided. Thus our objective was to

show the distinct functions of contempt, anger, and

disgust on ecologically valid data: contextualized

Spanish words for these three emotions. The first

part of this study was carried out on the Spanish of

Spain and had a descriptive objective; the second

part, on the Spanish of Latin America, had

comparison as its objective.

2 METHOD

2.1 Samples

This study was carried out on the Current Spanish

Reference Corpus, a stratified sample of

contextualized Spanish words from 1975 to 2004

(RAE, 2010). It is composed of more than

150,000,000 words and has been designed to offer a

representative sample of standard Spanish from

1975-2004. Criteria for text inclusion are clearly

specified: 50% from Spain, and 50% from America

(Mexican, Central, Caribbean, Andean, Chilean, and

River Plate area); 90% written, and 10% oral.

Various topics have been sampled: Science and

Technology, Social Sciences, Politics and

Economics, Arts, Health, Fiction, Leisure and

Everyday life. The first working corpus was

composed of every document (from Spain) including

the Spanish nouns for contempt, anger or disgust

(i.e., desprecio, rabia and asco). Note that this

corpus includes sayings and other expressions that

are not objectively comparable to heteroattributional

scenarios, in which emotions are predicated on

somebody else. Only 362 texts were finally selected:

the heteroattributional ones on whose coding two

“blind” observers had perfectly agreed. Of these,

some 141 texts corresponded to contempt, 124 to

anger and 97 to disgust. Our assumption was that

stringent criteria (i.e., perfect agreement on object

and attribution codes) would leave the noise in the

data out of the analysis, allowing the emergence of

clear association patterns. The second part of the

study was carried out on 139 texts from Latin

America: the heteroattributional ones on whose

coding two “blind” observers perfectly agreed. The

original working corpus for this comparative study

was composed of every document (from the three

largest Latin American Spanish-speaking zones)

including the Spanish nouns for contempt, anger or

disgust. Of these, 46 belonged to Argentina, 33 to

Chile and 60 to Mexico. It has been found that these

three Latin American countries differ in values from

Spain (Schwartz, 2008).

2.2 Procedure

The exhaustive and mutually exclusive category

systems were developed in two phases: (1) inductive

text categorization by a researcher blind to the CAD

hypothesis, and (2) deductive refining of the

categories to take into account theoretical codes

from previous research (Delgado, 2009a). Emotion

receivers and attributions were salient in most texts

and therefore these two structural elements

associated with social communication led to two

concurrent category systems. Finally two “blind”

observers, carefully selected for their excellent

grades (a procedure that warrants the high level of

reading comprehension that is needed for the task),

independently read and coded the texts following

instructions concerning the receiver and attribution

categories developed in the bottom-up part of this

study. Data were then quantitatively analyzed.

Various categories and subcategories were

eliminated from the analyses due to low frequencies

and/or agreement problems. The second part of the

study was carried out on the Spanish of Latin

America with the refined category systems. Two

new “blind” observers independently read and coded

the texts. Data were then quantitatively analyzed and

results compared with those from the Spanish of

Spain.

3 RESULTS

Results from the 362 selected texts indicate that, in

Spain, the typical emotion receiver is different for

contempt, anger and disgust scenarios,



(6)=

108.33, Cramer’s V = .39, p <.001. Standardized

residuals from Table 1 show that, while disgust takes

as its object mostly something concrete (e.g., feces,

bad breath), anger is preferentially felt towards

ICEIS2013-15thInternationalConferenceonEnterpriseInformationSystems

another person, and contempt towards an abstract

object such as terrorism or a social institution. The

category “oneself”, which was initially part of the

coding system, contained only seven cases and had

to be collapsed with the category “non-explicit or

other” given that



tests are not recommended when

expected frequencies are too low. In any case, the

CAD hypothesis proposes contempt, anger and

disgust as other-rejecting emotions.

Table 1: Number (corrected standardized residual) of

receiver categories by emotion in Spain.

Contempt Anger Disgust

Concrete object

(-5.3)

(-4.0)

(10.2)

Person

( 2.0)

( 2.9)

(-5.4)

Abstract object

( 2.2)

( -0.9)

(-1.5)

Non-explicit/ other

( -0.5)

( 1.3)

( -0.9)

Following current theorization on morality,

Table 2 attribution labels correspond to the coded

motives as follows: prejudice is used when the cause

of the emotion is “something negative that is

intrinsic to the object or person receiving the

emotion”; reciprocal is used when the cause of the

emotion is “something negative that the emotion

receiver has done to the person feeling the emotion”;

and the label altruistic is used when the cause of the

emotion is “something negative that the emotion

receiver has done to a third party”.

Table 2: Number (corrected standardized residual) of

attribution categories by emotion in Spain.

Contempt Anger Disgust

Prejudice

(1.0)

(-6.9)

(6.3)

Reciprocal

(-.9)

(4.4)

(-3.7)

Altruistic

(-4.4)

(6.8)

(-2.4)

Non-explicit/ other

(3.1)

(-0.6)

(-2.8)

Table 2 shows that the moral function of the

CAD emotions is present in our data: some

reciprocal attributions are found for anger, contempt

and disgust (in descending order), as well as

altruistic motives for anger and, in two texts, for

disgust. It can also be seen that significant

differences between contempt, anger and disgust can

be found concerning attributions,



(6)= 97.32,

Cramer’s V = .37, p <.001. Standardized residuals

indicate that disgust is positively associated with

prejudice, while anger is positively associated both

with reciprocity and altruism, the moral contexts. As

for contempt, there is only one large positive

adjusted standardized residual due to the twenty-

nine texts in the “non-explicit or other” category.

As to the comparative part of this study, results

from the 139 selected texts show that, in Latin

America, the typical emotion receiver is different for

contempt, anger and disgust scenarios (see Table 3),

closely replicating results from Spain,



(6)= 64,35,

Cramer’s V = .48, p <.001.

Table 3: Number (corrected standardized residual) of

receiver categories by emotion in Latin America.

Contempt Anger Disgust

Concrete object

(-1.9)

(-3.4)

(6.9)

Person

( -0.4)

(3.6)

(-4.1)

Abstract object

(3.7)

(-2.7)

(-1.3)

Non-explicit/ other

(-1.8)

(1.8)

(-0.1)

Table 4 shows that the moral function of the

CAD emotions is somewhat different in Latin

America: both reciprocal and altruistic attributions

are found for anger, contempt and disgust (in

descending order), and there is a significant

association between emotions and attributions,



(6)= 49,15, Cramer’s V = .42, p < .001, but residual

analyses show that contempt is associated with

prejudice, and that the altruistic function of anger

fails to reach significance. The association of disgust

with prejudice and anger with (lack of) reciprocity

did replicate results from Spain.

Table 4: Number (corrected standardized residual) of

attribution categories by emotion in Latin America.

Contempt Anger Disgust

Prejudice

(3.2)

(-6.4)

(4.3)

Reciprocal

(-2.3)

(5.1)

(-3.6)

Altruistic

(-1.4)

(1.9)

(-.6)

Non-explicit/ other

(-.2)

(.6)

(-.6)

4 CONCLUSIONS

In the descriptive part of this study, our analysis of

362 texts from a representative corpus of

contextualized words from Spain corroborates the

existence of a number of moral functions of

contempt, anger and disgust in everyday life, and

shows that both the emotion receiver and its

SocialRobots,Cross-culturalDifferences

perceived cause are different for these emotions:

disgust is mostly associated with concrete objects,

anger with persons, and contempt with abstract

objects.

Concerning attributions, disgust is mostly

associated with what we have denominated

prejudice (i.e., due to something negative that is

intrinsic to the object or person receiving the

emotion) and anger with what we have called

reciprocal and altruistic motives, while contempt

remains the most elusive of the emotion triad. The

moral role of contempt, the most salient of the CAD

emotions in Spain (Delgado, 2009b), was limited to

a number of reciprocal scenarios; the fact that there

are twenty-nine contempt texts in the “non-explicit

or other” category indicates that attributions for

contempt are the most subtle, and thus the most

difficult to categorize.

In the comparative part of the study, results from

Latin America replicated results from Spain

concerning the emotion receivers. The association of

disgust with prejudice and anger with (lack of)

reciprocity were again found. However, contrary to

expectations, the altruistic function of anger did not

reach significance and contempt was associated with

prejudice.

With respect to the CAD emotions, some cross-

cultural differences have already been reported:

Americans have been found to endorse contempt and

disgust expressions more often than Germans, who

endorsed anger more (Koopmann-Holm and

Matsumoto, 2011). Differences concerning the

moral functions of contempt in Spain and Latin

America go a step further by showing some

differences in meaning when language is the same.

Given differences in values between Latin American

countries and Spain (Schwartz, 2008), our results

could be explained by resorting to value-related

constructs. Our procedure is an ecologically valid

one that can be of help for designing more realistic

social robots.

ACKNOWLEDGEMENTS

This research was partially supported by research

grants MEC EXPLORA SEJ2007-29492-E and

MICINN PSI2009-09490. The authors wish to thank

Sergio de Dios and Angel Sanchez-Rodriguez for

acting as blind coders for the first part of this study

and Carlota Calvo and Marta Montero for acting as

blind coders for the second part.

REFERENCES

Bauer, M. W., Aarts, B., 2000. Corpus construction: a

principle for qualitative data collection. In M.W.

Bauer and G. Gaskell (Eds.), Qualitative researching

with text, image and sound. A practical handbook

(pp.19-37). London: Sage.

Baumard, N., Sperber, D., 2010. Weird people, yes, but

also weird experiments. Behavioral and Brain

Sciences, 33, 84-85.

Bloom, P., 2010. How do morals change? Nature, 464.

Cañamero, L., 2005. Emotion understanding from the

perspective of autonomous robots research. Neural

Networks, 18, 445-455.

Chapman, H. A., Kim, D. A., Susskind, J. M., Anderson,

A.K., 2009. In bad taste: evidence for the oral origins

of moral disgust. Science, 323, 1222-1226.

Cosmides, L., Tooby, J., 2005. Neurocognitive adaptations

designed for social exchange. In D. M. Buss (Ed.),

Evolutionary psychology Handbook (pp. 548-627).

New York: Wiley.

Delgado, A. R., 2009a. Social Robots, Moral Emotions.

Proceedings of the 11th International Conference on

Enterprise Information Systems (pp. 263-270).

INSTICC.

Delgado, A.R., 2009b. Spanish basic emotion words are

consistently ordered. Quality & Quantity, 43, 509-517.

Haidt, J., (2008). Morality. Perspectives on Psychological

Science, 3, 65-72.

Koopmann-Holm, B., Matsumoto, D., 2011. Values and

display rules for specific emotions. Journal of Cross-

Cultural Psychology, 42, 355-371.

Moll, J., de Oliveira-Souza, R., Zahn, R., 2008. The neural

basis of moral cognition. Sentiments, concepts, and

values. Annals of the New York Academy of Sciences,

1124, 161-180.

Real Academia Española, 2010. Spanish Royal Academy

Data Bank (CREA) [on-line]. Corpus de referencia

del español actual. <http://www.rae.es>

Rozin, P., Haidt, J., Fincher, K., 2009. From oral to moral.

Science, 323, 1179-1180

Rozin, P., Lowery, L., Imada, S., Haidt, J., 1999. The

CAD triad hypothesis: A mapping between three

moral emotions (contempt, anger, disgust) and three

moral codes (community, autonomy, divinity).

Journal of Personality and Social Psychology, 76,

574-586.

Schwartz, S. H., 2008. Cultural value orientations: Nature

and implications of national differences. Moscow:

State University Higher School of Economics Press.

Taylor, K., 2007. Disgust is a factor in extreme prejudice.

British Journal of Social Psychology, 46, 597-617.

ICEIS2013-15thInternationalConferenceonEnterpriseInformationSystems