unfuel

To deepen my understanding of the thesis topics, I conducted a literature review focused on three areas: persuasive technology (PT), sustainable mobility, and psychological behavior change theories. After defining key terms, I explored how these fields intersect and identified gaps my research could address.

PT, which aims to influence behavior without coercion or deception, is common in everyday life through tools like health trackers, road tachometers, and web ads. Extensive research offers various design strategies for PT, often grounded in behavior change theories, such as the theory of planned behavior and the transtheoretical model. For my work, I adopted the stage model of self-regulated behavior change (SSBC), specifically designed for sustainability contexts. The SSBC outlines four stages of behavior change and provides tailored intervention recommendations, which informed the messaging in my diary study. Additionally, research on PT in conversational agents helped shape an appropriate tone of voice.

The principal study spanned two weeks, during which participants received regular messages encouraging them to adopt or maintain sustainable transportation habits. Each day, they reported their mobility behavior, followed by a post-study interview to reflect on their experiences and the messages' impact. Before starting, a survey assessed participants’ current mobility habits and their willingness to change for sustainability. Based on the survey results, the ten participants were classified into one of the four SSBC stages of behavior change.

The next step was to design messages that fit the recommendations for interventions drawn out in the SSBC. Depending on the stage a participant was in, they would receive different messages. Due to only having participants in the predecisional and the preactional stage, it was only necessary to design messages for these two stages.

For the purpose of the study, I created an account on Telegram with an untouched telephone number, disguised as a bot to enhance the feeling of talking to an actual chatbot. I laid out a plan for the next two weeks at which day which message should be sent to which participant. Some of the messages had the explicit goal of targeting specific habits of certain participants for personalization purposes. These messages are italicized in the figure below.

After the diary study had concluded, I conducted interviews with the participants. The primary objective was to investigate the system’s persuasiveness, i.e. whether they had actually noticed a change in their mobility behavior. While some did not notice a change in the behavior itself, for others, this was actually the case. One participant noted that they had perceived the messages as a substantialmotivational factor in completing their daily commute via bicycle. Qualitative findings also revealed that many participants now thought much more about their mobility and its impact on the climate. Many behavioral change models regard this as a first step to actually changing a behavior.

Interviews and the diary study revealed that users are motivated by different factors, such as cost savings or health benefits. Research confirmed that persuasive technology is more effective when tailored to individual motivations. Therefore, the onboarding process included questions about these factors, allowing the chatbot to emphasize relevant aspects. A statistics section provided feedback on users' mobility behavior, with a simple indicator showing whether their goals were being met.

To validate the design, I conducted a UX test involving three tasks: completing the onboarding process, interacting with the chatbot, and retrospectively adding a trip to the archive. Participants shared their thought process to suggest improvements. The onboarding process was well-received, particularly the option to choose different communication styles. Users navigated the app with ease, with no major issues in information hierarchy. In the statistics section, all four participants requested a map showing their trips and preferred more detailed feedback than the traffic light scale, such as real-world comparisons. The app scored 83.8 on the System Usability Scale. In addition, the tests offered valuable insights into which directions future iterations of the applications could take.

My journey into the world of persuasive technology was interesting and, at times, challenging. Ultimately, I found that persuasive technology, particularly when paired with artificial intelligence, has a lot of potential to shape our behavior into ways that we feel are desirable. Key factors in maximizing efficiacy are personlization and a tailored address of the user, among others. In terms of design, it seems that the “hybrid” approach of leveraging AI not only in a chat interface, but in many other interaction spheres is the way to go moving forward.

As artificial intelligence becomes increasingly more integrated in existing technology, it seems that designers in key AI companies have come to the same conclusion. Whether or not all of these AI use cases actually bring value to the table remains to be seen.

The problem of non-sustainable mobility proved to be a complex. Repeatedly, it became clear during the work that it cannot be solved by individual action alone. Even if motivation is high, the infrastructure to support this willingness to change habits must exist. I advocate for an approach that accomodates both sides of the coin: Individual action as well as systemic change.