Female dogs evaluate levels of competence in humans

Reputation is considered to play an important role in maintaining humans’ large-scale cooperative societies. For example, someone’s reputation might influence people’s decisions about whether to behave altruistically toward that person or not. Even although an individual who engaged in altruistic behavior bears a temporary cost, if the receiver subsequently provides a selective advantage to the altruist, then so-called reciprocal altruism can evolve in the long run (Trivers, 1971). Reputations are formed when people make judgments about others based not only on their direct interactions with the latter, but also on information from interactions among third parties (Nowak, 2006, Wu et al., 2016). Making such judgments is variously referred to as “social evaluation”, “social eavesdropping” or “image scoring” (for a review, Abdai and Miklósi, 2016). It makes sense to prefer helpful individuals and avoid harmful ones when choosing a potential partner. There is some evidence for social evaluations in other species in which cooperation with conspecifics or allospecifics occurs (chimpanzees, Pan troglodytes (Subiaul et al., 2008); ravens, Corvus corax (Asakawa-Haas et al., 2016); cleaner fish, Labroides dimidiatus (Bshary and Grutter, 2006)). Recent studies have shown that some animals even socially evaluate abstract agents or simple moving shapes (bottlenose dolphins, Tursiops spp (Johnson et al., 2018); bonobos, Pan paniscus (Krupenye and Hare, 2018)). Some non-human primates in captivity also socially evaluate humans (chimpanzees: Russell et al., 2008, Subiaul et al., 2008; tufted capuchin monkeys, Sapajus apella: Anderson et al., 2013a, Anderson et al., 2013b; common marmosets, Callithrix jacchus: Kawai et al. (2014)), even though that specific ability is unlikely to be required in the wild.

Compared to primates, domestic dogs (Canis familiaris) have adapted to living in human society through an evolutionary process (Miklósi et al., 2004). Dogs were the first domesticated animal, starting from around 15,000 years ago (Savolainen et al., 2002). During this long process of co-existence with humans, dogs have acquired relatively sophisticated abilities to read human behavior and communicate with humans (Hare et al., 2002, Hare and Tomasello, 2005). They also have innumerable opportunities to learn about human behavior through direct and indirect interactions with humans, making it conceivable that dogs socially evaluate humans on a daily basis. These considerations make dogs a highly suitable nonhuman species in which to study “eavesdropping” on humans (Freidin et al., 2013, McAuliffe et al., 2019).

Several studies have examined whether dogs can use their own direct experiences to distinguish between humans. For example, dogs learned to prefer a “generous” experimenter who allowed the dog to eat food rather than a “selfish” experimenter who ate the food before the dog could have it (Carballo et al., 2015). Such preferences are not limited to situations involving food. Nitzschner et al. (2012) used social reinforcement instead of food, and showed that dogs preferred an experimenter who behaved in friendly fashion toward them over one who ignored them. These results indicate that dogs prefer someone associated with beneficial outcomes (i.e., food or social treats).

Dogs are also sensitive to third-party interactions. Kundey et al. (2011) and Marshall-Pescini et al. (2011) tested dogs in a food sharing context. Although in these studies dogs preferred a “generous” experimenter who gave food to a begging person over a “selfish” experimenter who withheld food, it is possible that dogs simply remembered the one who was previously associated with food. To address that concern, in a previous study we tested dogs in a helping context that did not involve food (Chijiiwa et al., 2015). Dogs watched as their owner tried to open a container to remove an object inside the container, but after failing to open it the owner asked for help from one of two people sitting beside her. In the Helper condition the actor assisted the owner, whereas in the Non-helper condition she refused to help. The other person simply remained still and neutral throughout these interactions. After the interaction, the actor and the neutral person each offered a piece of food to the dog. The finding was that dogs avoided the non-helper in favor of the neutral person, whereas they did not show any preference for the helper vs. the neutral person. This asymmetrical preference reflected a “negativity bias,” which is the avoidance of a negative (mean) partner, reported in human infants by Hamlin et al. (2010) and in tufted capuchin monkeys by Anderson et al., 2013a, Anderson et al., 2013b.

In contrast, Nitzschner et al. (2012) found that dogs evaluate based only on direct experience; dogs did not prefer actors who were friendly to other dogs compared to actors who ignored them. Nitzschner et al. (2014) suggested the possibility that dogs merely choose a location rather than a human, since they preferred a “generous” experimenter over a “selfish” one only when the experimenters were not replaced before the dogs were given choice. Thus, findings regarding dogs’ ability to evaluate humans based on indirect experiences are inconclusive.

In humans, “competence”, which includes characteristics such as effectiveness, skills, creativity, confidence, and intelligence is one important dimension on which social evaluations are made (Cuddy et al., 2008; Sternberg, 2013). It could be advantageous for social animals to evaluate others based on not only social traits such as cooperativeness, but also on nonsocial traits such as skillfulness or competence, for example in observational learning contexts. Some primates show recognition of other conspecifics’ skills in motor tasks, for example by being more affiliative toward “specialist” individuals the group (Stammbach, 1988), paying attention especially to skilled models (Ottoni, 2005), and choosing to cooperate with previously successful individuals (Melis et al., 2006).

However, it is unclear whether dogs can discriminate humans based on the latters’ skillfulness. Horn et al. (2012) reported that dogs spent more time near a person whose role was to replenish a food-delivering apparatus when it was empty, but they approached another, helpful person first, irrespective of whether the apparatus was blocked or empty. Those authors also pointed out possible limitations to dogs’ understanding of the problem and its solution. In Piotti et al. (2017), dogs observed two types of demonstrations: a skilled experimenter successfully opening a special apparatus and getting food for the dog, and an unskilled experimenter whose attempts failed. To retrieve the food, the skilled experimenter turned the bottle upside down and pulled out a wooden partition. In the test phase, the dogs faced an unsolvable task in which the food was inaccessible to them; the two experimenters (skilled and unskilled) stood one on either side of the apparatus. In this situation, the dogs did not request help by looking back especially at the skilled experimenter. Again, in this study the apparatus might have been too complicated for the dogs; Piotti et al. (2017) pointed out the potential importance of simplicity and conspicuousness of manipulations when testing dogs’ understanding of competence.

Here, we investigated dogs’ recognition of human competence by using a simpler situation than in previous studies. We showed dogs demonstrations by two actors: one managed to remove an object from a container by opening the lid (Competent person), whereas the other’s attempts to open the lid failed (Incompetent person). The demonstration phase was followed by a test in which the two actors simultaneously tried to open a new container which either had food (Food condition), or nothing inside (Empty condition). We used a neutral object instead of food in the demonstration to avoid formation of any prior association between food and actor. We predicted that if dogs were sensitive to differences in competence, in the Food condition they would preferentially approach the Competent person, to maximize their chance of getting food. By contrast, in the Empty condition dogs would show no clear preference between the two actors, as there was no reason to anticipate getting food from either one.