Siegel, M., & Breazeal, C., & Norton, M., (2009). Persuasive Robots: The Influence of Robot Gender on Human Behavior.

From Control Systems Technology Group
Jump to navigation Jump to search

Persuasion is a fundamental aspect of human social interaction.
It’s needed for example when a search-and-rescue robot wants disaster victims to follow important instructions, a weight-loss-coach robot wants to assist people in changing their diet and exercise habits.
A better understanding of this area also reduces the chance robots being unintentionally designed to manipulate or influence humans in unexpected or negative ways.
In human-human interaction, many factors influence persuasiveness. Some examples: appearance, style, content of communication and non-verbal behavior.
In this study the role the gender of the robot has is examined.


The gender of the robot is determined by the use of a pre-recorded masculine or feminine voice. The experiment considers all combinations of the following: 2 robot genres (male/female), 2 subject genders (male/female), 2 situations (subject alone/subject not alone), resulting in 8 situations.


The participants were situated in a distract free area with the robot. They were given a $5 compensation beforehand for participating in the study. They were told that the robot may ask for a donation and it was their choice to give any of the money away.


The robot first provided a brief explanation of its hardware and software systems and gave a general overview of its technical capabilities. Then the robot presented a persuasive appeal arguing that the “uneven distribution of technology is one of the most important issues facing our world today.” The appeal ends with the following donation request. After donations were placed the robot asked subjects to fill out a short questionnaire.


In the questionnaire the following attitude measures were asked: trust, credibility and engagement. Trust was measured using a standard fifteen question scale. Credibility was categorized into safety, dynamism and qualification. Engagement is from Lombard and Ditton’s scales measuring the six aspects of presence.


The results for the donation measure did not follow a normal distribution, but peaked at $0 and $5. To simplify the analysis, this was treated as binary: gave nothing vs gave something. The result was that subjects donated more often to the female robot than the make robot (83% vs 56%). Also it was revealed this effect was primarily attributed to men, while women did not show a statistically significant preference for a specific gender.
Incorporating whether the subject was alone or not showed that men donated similar in both situations, but women donated significantly more often to the female robot when accompanied, but when alone they preferred donating to the male robot, though not significantly.


Whether the subject was alone or not did not have significant influence on the results on the questionnaire. So this attribute was dropped for the analysis.
Credibility did not exhibit a main effect for robot gender or subject gender. Men rated the female robot more credible than the male robot (78.8% vs 73.44%), women rated the male robot as more credible (80.8% vs 75.2%).
Trust also showed no significant main effect for robot or subject gender, but did exhibit the same cross-gender interaction effect: men found the female robot more trustworthy (82.8% vs 74.68%) and women found the male robot to be more trustworthy (81.62% vs 77.37%).
Splitting the cases revealed that men were predominantly affected by the change in robot gender, while women showed very little preference.
Also engagement showed the same effect: men reported being more engaged to the female robot (23.56 vs 17.28) and women were more engaged to the male robot (23 vs 22.81). Unlike others, engagement does show a mean effect with both independent variables. The female robot is more engaging than the male robot (23.26 vs 19.89) and women tend to report being more engaged than men (22.89 vs 21.18).
As with the trust measure, separating the cases based on subject gender shows that men reported significantly more engagement with the female robot, while women showed no preference.


Interesting addition: A recent study by Schermerhorn et al. gives some reason to believe that men will more readily treat a robot as a social entity, while women viewed a robot as more machinelike and did not show evidence of social facilitation (the propensity for people’s performance on certain tasks to change when being observed).