Reciprocity of social influence: Who do we take advice from and why?
Experiment set up
Every day we have to make lots of decisions and we try to optimize them. An important factor of optimization is to integrate advice from others. But who do we take advice from, why and how much? “Common sense would tell us: From a person we believe to be competent,” says Carsten Mehring, Professor of Neurobiology and Neurotechnology and Co-Director of the Bernstein Center Freiburg. Several studies confirm this view. However, he, his PhD-student Ali Mahmoodi and Bahador Bahrami from University College London have now carried out a study that can't be explained by that.
Their findings show: “Beyond the competences of a person: You take more advice from a partner who takes more advice from you and you take less advice from a partner who takes less advice from you,” Ali Mahmoodi explains. This means that participants reciprocated influence with their partner.
Since social interaction between people in general is often mutual, it may be close at hand that reciprocity also takes place when we receive and impose influence on others. “But as far as we know, this question has never been investigated up until now,” says Ali Mahmoodi. “Our results now prove that the amount of advice we take highly depends on how much our partner is willing to take advice from us. So our influence on others highly affects how much we are influenced by them.” And Carsten Mehring adds: “The participants’ behavior is not explained by their impression that one partner has a higher competence than the other – because the partners’ performance was always the same.”
Interestingly, the scientists have been able to show another astonishing effect: The participants in the study only reciprocated with an alleged human partner. If, on the other hand, they assumed they were interacting with a computer, reciprocity disappeared. This result came out even though the partner was a computer in every experiment, hence it only depends on your belief that your partner is a human.
Last but not least, the researchers were able to show a third interesting point: The participants’ estimates of their own performance massively dropped if they worked with a partner who was not influenced by the participants. The conclusion is that reciprocity also has a profound impact on the assessment of oneself.
Interpretation
“Our hypothesis was that people would reciprocate influence with others because reciprocity is a pervasive social norm,” says Ali Mahmoodi. And Bahador Bahrami, collaborator in this study from University College London, suggests: “We interpret this reciprocal behavior as an important communication tool in which group members communicate with their peers. Our findings point to an important difference between human-human and human-computer interactions. This difference helps anticipate possible emergent problems that may arise in future as AI agents are integrated in everyday human interactions.”
Experimental design
To find evidence for their hypotheses the scientists designed a behavioral experiment. Test persons played with unknown, invisible computer partners which they thought to be humans. The probands got to see a dot on a computer screen that moved quickly. They had to estimate where they perceived the dot first. The partner was given the same task. Then the test persons shared their estimation with their alleged partners and learned how their partners had estimated. In alternating trials, the probands either got the possibility to revise their estimates or observed how the partners revised theirs.
The researchers systematically manipulated the influence from the participants on the partners’ decisions. An algorithm simulated susceptible partners, who took advice from the probands, or insusceptible partners, who did not take advice from the probands. Over time, it became clear that the more often the partner followed the test persons’ suggestion, the closer the test persons’ revised decision came to that of the partners. The scientists then carried out further experiments. Participants thought they were playing either with a human or a computer. The result: participants only reciprocated with their partners if they believed that their partner was human.
Outlook
In the next step the researchers are seeking to get a better understanding of the underlying neuronal mechanisms of the observed behavior. “Currently we are doing similar experiments with people in a brain scanner and at the same time we are looking at what kind of brain activities we can observe,” Ali Mahmoodi says.
Original publication
Mahmoodi A.; Bahrami B. & Mehring, C. (2018) Reciprocity of social influence. Nature Communications 9, Article number: 2474
Contact
Ali Mahmoodi
Bernstein Center Freiburg
Hansastraße 9a
79104 Freiburg i.Br.
Tel.: +49 (0)761 203-9595
E-mail: ali.mahmoodi@bcf.uni-freiburg.de