«Naturalizing Anthropomorphism: Behavioral Prompts to Our Humanizing of Animals Alexandra C. Horowitz* and Marc Bekoff† *Department of Psychology, ...»
ANTHROZOÖS VOLUME 20, ISSUE 1 REPRINTS AVAILABLE PHOTOCOPYING © ISAZ 2007
PP 23 – 35 DIRECTLY FROM PERMITTED PRINTED IN THE UK
THE PUBLISHERS BY LICENSING ONLY
Behavioral Prompts to Our Humanizing of Animals Alexandra C. Horowitz* and Marc Bekoff† *Department of Psychology, Barnard College, New York, USA †Department of Ecology and Evolutionary Biology, University of Colorado, Boulder, USA
Anthropomorphism is the use of human characteristics to deAddress for correspondence:
Dr. Horowitz, scribe or explain nonhuman animals. In the present paper, we propose a Barnard College, Department model for a unified study of such anthropomorphizing. We bring together of Psychology, 353 Riverside Drive, previously disparate accounts of why and how we anthropomorphize and New York, NY 10025, USA
suggest a means to analyze anthropomorphizing behavior itself. We introE-mail:
duce an analysis of bouts of dyadic play between humans and a heavily firstname.lastname@example.org thropomorphized animal, the domestic dog. Four distinct patterns of social interaction recur in successful dog–human play: directed responses by one player to the other, indications of intent, mutual behaviors, and contingent activity. These findings serve as a preliminary answer to the question, “What behaviors prompt anthropomorphisms?” An analysis of anthropomorphizing is potentially useful in establishing a scientific basis for this behavior, in explaining its endurance, in the design of “lifelike” robots, and in the analysis of human interaction. Finally, the relevance of this developing scientific area to contemporary debates about anthropomorphizing behavior is discussed.
Keywords: anthropomorphism, attention, cognitive ethology, dogs, humanizing animals, social play ❖
Significantly, anthropomorphism, despite its accepted status as a fundamentally flawed way to describe nonhuman animal behavior, has long endured, both within the scientific community and especially in the general public’s conception of animals. Among lay people, anthropomorphism is not only prevalent, it is the nearly exclusive method for describing, explaining, and predicting animal behavior—whether the animals are kept as pets, visited in the zoo, or observed in nature.
The strong tendency to use human terms to describe animal behavior has prompted some to ask why we anthropomorphize at all (Caporael and Heyes 1997). The behavior and causes of anthropomorphizing have not been the subject of systematic analysis (Guthrie 1997). In the present paper, we begin to remedy this with an initial analysis of the behaviors of social play between dogs and humans. In particular, we ask: What behaviors by dogs in play correspond to the behavior of projective anthropomorphizing by owners? By looking at the context of this class of anthropomorphisms, the relevant attributional stances can be reduced to a few behavioral cues. Such an analysis may be informative not only about the subjects of the attributions, but also about the attributers.
In this paper, we begin with the precedents to a study of anthropomorphizing; then we apply its method to behavioral data of dog–human play. We propose to use ethology as a model for the study of the behavior. To this analysis we add prior accounts of the prompts for anthropomorphism from various sources: robotics, social science, and ethology. Finally, we consider the relevance of an analysis of anthropomorphizing to traditional discussions of anthropomorphisms in science.
An Ethological Approach to Anthropomorphism The science of ethology provides a methodology for considering anthropomorphizing behavior analytically. Ethology is the study of animal behavior in natural settings, using live observations and videotapes to record data of focal behaviors or subjects, transcription of the recorded events in an inventory of behavioral units coded per an ethogram, and analysis of behavior patterns. We propose that data of use of anthropomorphisms will show that they are reliable and identifiable enough that they may themselves be treated ethologically. Just as the observational method of ethology has provided a means to naturalize sociality, aggression, communication, theory of mind, and many other phenomena by tying these concepts to seen behaviors, the attributions humans make to animals can be characterized and correlated with seen behaviors through ethological methods.
By examining both general and specific application of anthropomorphisms, a catalogue of physical and behavioral features which lead to anthropomorphizing behavior in humans could be formed.
This method may yield results with application to the design of lifelike robots, and lead to a clearer understanding of humans’—scientist and lay alike—tendency to anthropomorphize.
Theoretical precedent for this approach comes from the comparative psychologist Donald Hebb. In his time at Yerkes, he noted that in contrast to his own non-attributive description of the captive chimpanzees’ behaviors, the keepers’ anthropomorphic accounts provided a useful guide to the behavior of their charges (Hebb 1946). Rather than working to eliminate anthropomorphisms outright, Hebb and the other researchers recorded the behavior patterns of the animals alongside their anthropomorphic glosses. In particular, they specified those behavior patterns that led to the attribution of emotions such as fear, nervousness, and shyness. This yielded a psychological description of animal behavior (Mitchell and Hamm 1996).
Even though Hebb did not see it as a study of anthropomorphism per se, his results support the theory that anthropomorphizing is a well-defined class of behaviors that has clear manifestations, Anthrozoös clear triggers, and evaluative usefulness. We have appropriated and refined Hebb’s method for our present ethological analysis. Such analyses seem to be gaining momentum, as presaged by Mitchell and Thompson’s (1986a) discussion of the viability of investigating which “biological and psychological processes” lead to the attribution of deception in nonhuman animals. Three studies have looked at subjects’ reports of what is “going on” in a film playback: of moving shapes (Heider and Simmel 1944; Berry et al. 1992) and of dogs interacting with their caregivers (Morris, Fidler and
anthropomorphisms; Morris, Fidler and Costall found that subject reports were consistent, anthropomorphic, and were reliably matched to episodes of behavior, distinct “event units.” We present a complementary approach: identifying the repeated behavior patterns in just those bouts of play which the (human) participants identified as “successful” or “satisfying,” and expressed that their dogs understood or enjoyed the play.
Anthropomorphisms in Dog–Human Play In just what circumstances are anthropomorphisms made? We use the ethological data of interaction between the anthropomorphizers—humans—and the anthropomorphized—dogs—to discover regularities in the human attributions. The result is a preliminary catalogue of the behaviors that correspond to one kind of anthropomorphizing—the attribution of playful intent or understanding—by humans engaged in dyadic play with their dogs. In this paper, we focus on the proximal causes of anthropomorphism; further research might explore the evolutionary or predictive uses thereof.
The data from which this analysis draws are the result of research done to advance the design of a pet-like robot, Sony’s “Aibo.” In 2003 one of us (A.H.) performed an observational study of dog–human interactions in play to create a catalogue of play behaviors (more details of the methods of this study are available elsewhere1). One goal in the design of the Aibo robot is to “encourage human and robot interaction” (Sony entertainment robot Aibo operating instructions 2000). The robot is dog-like in its form—it is four-legged, has a tail, characteristic head form, etcetera—and behavior—it ambles and wags like a dog, barks, and performs simple trained-dog routines—and is intended to interact with human beings. The research was motivated by a desire to create play routines and games which could be modeled in a dog-like robot such as Aibo. With this method, what initially appear to be highly complex social interactions can be deconstructed, and then reconstructed in the robot.
The data also capture a variety of the ways that humans interact with, and the expectations they have of, a creature which holds special status in American culture: the domestic dog. Dogs are distinguished not only by being one of the only animals granted access inside the homes and families of a majority of the population (U.S. Pet Ownership Statistics 2006)—they also are particularly subjected to our tendency to anthropomorphize. The plethora of books published annually on the subject of dog-ownership tread easily and without controversy on the topics of peoples’ “relationships” with dogs, the animals’ “understanding” and “love;” what dogs “know,” “like,” “think about;” dogs who are “shy,” “trying to tell you” something, who have a “point of view;” who experience “grief” and “friendship.” They are fitting subjects for a new examination of why we anthropomorphize.
This study of dog–human play behavior is part of the growing field of study of dog–human interaction (Sanders 1999; Podberscek, Paul and Serpell 2000; Irvine 2004) and of cognitive phenomena in dog–human play in particular (Mitchell and Thompson 1986b; Mitchell and Thompson 1991; Rooney, Bradshaw and Robinson 2001). Following the ethological methods of these and other studies of social play (e.g., Tomasello et al. 1994; Bekoff and Byers 1998; Horowitz 2002), in this research videos were taken of humans interacting in a self-described playful way with their own dogs. For these purposes, “play”—which, though easy to recognize, is more difficult to characterize (Bekoff and Byers 1998)—was defined as voluntary, coordinated behavior between a human and a dog which often followed a routine form (characterized below). A play bout was considered complete if the episode included identified play behaviors (Horowitz 2002) and lasted for two or more turns on each participant’s part. The “routines” (Mitchell and Thompson 1986b) seen Anthrozoös included object retrieval games (fetch: throw, retrieval, and sometimes pursuit); object possession games (tug-of-war; keep-away); social playmate games (chase); feigning games (wrestling; growling; vocalizing); and parallel behavior (running alongside one another; mutual posture-changing).
Play occurred naturally, and was not manipulated by the observer. Play bouts took place and were recorded in public settings—the majority in Central Park, New York City, and enclosed parks des
ignated “dog parks” in the city. Naturalizing Anthropomorphism
The data from this study can be appropriated to consider the behavioral correlates of anthropomorphism. In post-play questionnaires and in volunteered concurrent commentary, the human players often projected psychological states onto the animals: their dogs “understood” the game and “know” what they were “supposed to do;” the dogs were playing “fair,” being “cooperative” or “sneaky,” or were trying to “trick” their partner; their dogs “like” playing with them; they were “having fun,” and “loving it.” These descriptions occurred during or after “successful” play, as per the owners’ assertions that the play had gone well. They were thereby distinguished from those in which the human players expressed frustration or abandoned the bout prematurely. Fifteen (15) independent “successful” sessions of dogs interacting with their human caretakers in object retrieval, object possession, and feigning games were selected for analysis. Insofar as the estimations of the dogs’ psychological states are unproven (by the owners and by behavioral scientists), they can fairly be called “anthropomorphisms.” The analyzed bouts serve as the medium for anthropomorphisms of playful intention or understanding.
Examination of the ethological record of behaviors of these play bouts, supplemented with data from behavioral analysis of dyadic dog play (Horowitz 2002), reveals distinctive patterns in the interaction between players.
Categories of Social Behavior From the catalogue of dog and human behaviors across play bouts, we have identified numerous behaviors that repeat in successful bouts—and that are often absent in unsuccessful bouts or when play collapses. These examples suggest a grouping based on their commonalties. Four “social categories”—distinct groups of behaviors or patterns of behavior which expose and represent a feature of social interaction—are introduced below. Capsule summaries of some of the exemplar behavior patterns noted across play sessions follow.
(1) Directed responses to the other: A characteristic phenomenon of play bouts is each player’s regular and reliable responsiveness to the other player: a sort of verification of the other’s participation in the dialogue. This is particularly manifested by the use of the attention of the other player. A coordinated use of attention appears to be integral to play (Tomasello et al. 1994; Horowitz 2002).
Every successful dog–human bout had some example of this. For instance, in this study, games often began with full-face contact between the human and the dog: each oriented his face to the other’s face. In some instances, this contact was additionally seen as the game unfolded. In a typical game of fetch, for example, every time the ball was returned to the owner, the human player paused until they shared eye or face contact before throwing the ball again. In the same way, by orienting one’s face away from the other’s, a play session may be ended.