We are delighted to present this special issue of Behaviour & Information Technology (BIT), featuring extended versions of a selection of the best papers presented at the 18th International Confere...

This article explores the relation between consistency of social cues and persuasion by an artificial agent. Including (minimal) social cues in Persuasive Technology (PT) increases the probability that people attribute human-like characteristics to that technology, which in turn can make that technology more persuasive (see, e.g., Nass, Steuer, Tauber, & Reeder, 1993). PT in the social actor role can be equipped with a variety of social cues to create opportunities for applying social influence strategies (for an overview, see Fogg, 2003). However, multiple social cues may not always be perceived as being consistent, which could decrease their perceived human-likeness and their persuasiveness. In the current article, we investigate the relation between consistency of social cues and persuasion by an artificial agent. Findings of two studies show that consistency of social cues increases people’s recognition and recall of artificial agents’ emotional expressions, and make those agents more persuasive. These findings show the importance of the combined meaning of social cues in the design of persuasive artificial agents.

Earlier research showed that artificial social agents can influence human behavior. This article argues that, especially under certain circumstances, people are sensitive to persuasion by (artificial) social agents. For example, when people feel socially excluded, they are motivated to increase their social connections with others. It was hypothesized that socially excluded people would attribute more human-likeness and be persuaded more by an artificial agent than socially included people. These hypotheses were investigated in two studies in which participants were either socially included or excluded, after which they performed an energy-saving task while receiving social feedback from an artificial agent. Results did not support the expectation that socially excluded people ascribe more human-likeness to an artificial agent, but they did show the expected effects on behavior change, indicating the importance of including a person’s psychological state in the design of human–agent interactions. Also, in line with earlier findings, female participants were more susceptible to the agent’s feedback than male participants, indicating that a user’s gender may also determine the effectiveness of persuasive technology.

In human-robot interaction research, much attention is given to the extent to which people perceive humanlike attributes in robots. Generally, the concept anthropomorphism is used to describe this process. Anthropomorphism is defined in different ways, with much focus on either typical human attributes or uniquely human attributes. This difference has caused different measurement tools to be developed. We argue that anthropomorphism can best be described as a continuum ranging from low to high human likeness, and should be measured accordingly. We found that anthropomorphic characteristics can be invariantly ordered according to the ease with which these can be ascribed to robots.

In this paper we argue that energy conservation is largely a group phenomenon requiring group interventions to achieve change. Persuasive technology can help to provide these interventions. The present study explores the influence of group feedback and individual comparative feedback on energy consumption using an experimental simulation paradigm. To account for cultural differences in group orientation and the power of group feedback, two studies were conducted, one in the Netherlands and one in Japan, in which groups of participants received feedback on everyday tasks. As expected, Dutch participants saved more energy when individual comparison feedback was present, but not the Japanese participants. In contrast, as expected, group feedback caused Japanese participants to save more energy. Providing solely group feedback did not promote energy saving in the Netherlands. Group feedback made the Dutch save more energy only in combination with individual comparison feedback. These results suggest that persuasive technology can employ the power of feedback as a group intervention, but that relevant cultural orientations are crucial.

A fundamental characteristic of ambient persuasive technology is the ability to persuade users outside of conscious attention. One method of influencing people outside of their conscious awareness is subliminal priming: Presenting a stimulus for less than 50 milliseconds so people can not consciously perceive it but they are able to process the information unconsciously. Earlier research has shown that subliminal feedback on energy consumption leads to more optimal choices in an energy-related choice task compared with no feedback. Would subliminal feedback always be effective in influencing people's choices, or do people need to be motivated to reach a specific goal for it to work? In the current research, we investigate if persuasion by subliminal feedback comprises a goal-striving related process in which people strive for a goal and use the subliminal information to reach that goal. In an experimental study, half of the participants were primed with the goal to perform well, and the other half was not primed with this goal. Next, half of the participants received subliminal feedback in a learning task, whereas the other half received no feedback. Results indicated that participants primed with a performance goal made more correct choices in the task when given the subliminal feedback compared with given no feedback. Participants who were not primed with a goal were not influenced by this feedback. This finding indicates the need for a behavior-relevant goal to make subliminal information effective, which in turn has important implications for research and design of ambient persuasive technology.

Two experiments explored the persuasive effects of social feedback, as provided by a robotic agent, on behavioral change. Results indicate stronger persuasive effects of social feedback than of factual feedback (Experiment 1) or factual evaluative feedback (Experiment 2), and of negative feedback (especially social but also factual) than of positive feedback.

In this paper we explore a fundamental characteristic of Ambient Persuasive Technology: Can it persuade the user without receiving the user's conscious attention? In a task consisting of 90 trials, participants had to indicate which of three household appliances uses the lowest average amount of energy. After each choice, participants in the supraliminal feedback condition received feedback about the correctness of their choice through presentation of a smiling or a sad face for 150 ms. Participants in the subliminal feedback condition received identical feedback, but the faces were presented only for 25 ms, which prohibited conscious perception of these stimuli. The final third of the participants received no feedback. In the next task, participants rated the energy consumption of all presented appliances. Results indicated that supraliminal feedback and subliminal feedback both led to more correct energy consumption ratings as compared to receiving no feedback. Implications are discussed.

In this paper we explore the persuasive effects of social feedback, as provided by an embodied agent, on behavioral change. In a lab setting, two experiments were conducted in which participants had the opportunity to conserve energy while carrying out washing tasks with a simulated washing machine. The experiments tested the effect of positive and negative social feedback and compared these effects to more widely used factual feedback. Results of both studies indicate that social feedback has stronger persuasive effects than factual feedback (Experiment 1) and factual-evaluative feedback (Experiment 2). In addition, an effect of feedback valence was found, demonstrating more conservation actions following negative feedback (social or factual) as compared to positive feedback. Interestingly, especially negative social feedback had the strongest persuasive effects. The predicted perceived agency effect could not be demonstrated. These findings have several implications for theory and design of persuasive robotic agents.

Nowadays, many advertising campaigns attempt to persuade people to perform a specific behavior. In response to such messages, people can comply and adapt their behavior in the proposed direction. However, people can also experience psychological reactance, which may lead to the complete opposite of the target behavior. In the present study, we were interested in the social nature of psychological reactance. According to Social Agency Theory [12], more social cues lead to more social interaction. We suggest that this also holds for psychological reactance. We argue that there is a positive relation between the level of social agency of the source of a message and the level of psychological reactance that this message can arouse. In an online study, participants received low-controlling or high-controlling advice about energy conservation. This advice was delivered either solely as text, as text with a still picture of a robotic agent, or as text with a brief film clip of the same robotic agent. Results showed that a high-controlling advisory message resulted in more reactance than a low-controlling advisory message. Confirming our expectancies, stronger social agency of the messenger led to more psychological reactance. Implications are discussed.

Feedback can serve as an intervention aimed at reducing household energy consumption. The present study focused on the effects of agent embodiment on behavioral change through feedback. The effects of agent embodiment were studied for female vs. male users. Also factual feedback was compared to evaluative feedback. An experiment was conducted in which 76 participants used a virtual washing machine to clean laundry. They received interactive feedback about their energy consumption, from an embodied agent or from a computer. This feedback indicated the consumption level (factual feedback) or good or bad performance (evaluative feedback). The results showed that evaluative feedback, especially when it was negative, was more effective than factual feedback in reducing energy consumption, independent of the source of the feedback. Overall, for men it did not matter whether the feedback was given by a computer or by an embodied agent, but for women it did: women who interacted with the embodied agent used less energy than women who interacted with the computer.

In contrast with most other experimental system trust research, this paper examines indirect information as a basis for trust. In experiment 1, the overall valence of an evaluation concerning a route planner was pitted against a consensus cue, i.e. a favourable opinion about the system endorsed by a minority versus a majority. A positive evaluation caused an increase of system trust, whereas a negative evaluation led to a decrease. Control allocation, i.e. choosing manual or automatic mode, however, remained unaffected. Furthermore, no effect was found of consensus; one explanation holds that, despite the absence of outcome feedback, displaying of routes on-screen provided interfering trust-relevant information. Focusing solely on the consensus effect in the absence of route display, experiment 2 revealed consensus to affect both trust and control allocation. These experiments show that trust-relevant information can be processed heuristically and systematically. Possibly, trust can also be based on direct information despite absence of feedback whether generated solutions are good or bad.

In general, users act as if their behavior is controlled by a self-regulatory system (Carver & Scheier, 1998) during user-system interaction. The self-regulatory system is specified as a feedback control mechanism consisting of monitoring, evaluation, and action adjustment mechanisms. This article investigates whether perception processes take care of monitoring, and if a user's evaluation is based on the interpretation of the results of such monitoring. To do this, a typical system property, time delay, was manipulated. In an experiment, the relation between estimated task duration and actual time-to-task completion was studied. The findings support the premise that perception processes, interpreted as a monitoring mechanism, can keep track of physical task time. A monotonic, positive relation was found between the time-to-task completion and user satisfaction, which supports the assumption that objective efficiency influences the evaluation of interaction. A linear relation between the estimated duration and user satisfaction indicates that the monitoring mechanism transforms physical signals into a format that can be used in the evaluation process, thereby changing the internal reference frame of the evaluation mechanism.