TL;DR
This study applies Self-Determination Theory (SDT) and automation trust to explain why users intend to adopt new interaction technologies in autonomous vehicles (AVs). The researchers found that psychological needs—specifically autonomy and trust—are universal drivers of adoption, but the importance of "relatedness" and "competence" varies significantly depending on whether the technology is a functional alert system or a social virtual assistant.
What Problem Does This Paper Address?
As autonomous vehicles (AVs) evolve, they are shifting from being mere "tools" to becoming "intelligent assistants." While traditional research has focused heavily on the Technology Acceptance Model (TAM)—which emphasizes extrinsic factors like perceived usefulness and ease of use—there is a gap in understanding the intrinsic psychological motivations that drive long-term adoption.
The authors tackle a specific challenge: How do our basic psychological needs (the desire to feel in control, capable, and connected) interact with our trust in automation to influence our willingness to use new AV features? Crucially, they investigate whether these motivations change across different types of interaction media, such as functional windshield alerts versus anthropomorphic virtual assistants.
Methods
The researchers employed a within-subjects experimental design with 155 licensed drivers in Beijing.
- Stimuli: Participants watched high-quality introductory videos of two distinct AV technologies:
- Interaction Mode: A functional "Pedestrian Crossing Alert" that displays timing data on the windshield.
- Virtual Image: An anthropomorphic virtual assistant (from the Baidu Apollo platform) that uses eye contact and movement to communicate.
- Measurement: After each video, participants completed validated scales measuring:
- Self-Determination Theory (SDT) components: Competence, Autonomy, and Relatedness.
- Automation Trust: Using the 12-item Jian et al. scale.
- Behavioral Intention: The likelihood of using the technology.
- Analysis: The team used hierarchical multiple linear regression and bootstrap mediation analysis (via AMOS 21.0) to determine how psychological needs predict intention and how trust mediates that relationship.
Datasets
The study relied on a primary dataset collected specifically for this research:
- Sample: 155 drivers (87 male, 68 female).
- Demographics: Aged 21–40 (Mean = 27.93), all with at least one year of driving experience in Beijing.
- Stimuli Source: The interaction scenarios were adapted from Baidu Apollo autonomous vehicle introduction videos.
Key Insights and Ideas
- Autonomy is Non-Negotiable: Across both types of technology, perceived autonomy was a core predictor of intention. Users don't want the car to take over entirely; they want the "decision opportunity." Designers who automate everything (like air conditioning or music) without user input may actually decrease the intention to use those systems.
- The "Relatedness" Paradox: Interestingly, Relatedness (the feeling of being connected to others/the system) was a significant predictor for the functional Interaction Mode but not for the Virtual Image. This suggests that users view functional alerts as a way to stay connected to the driving environment, whereas a virtual assistant might be seen more as a "gimmick" that doesn't yet satisfy a deep need for connection.
- Trust as a Bridge: Automation trust acts as a partial mediator. Satisfying a user's psychological needs (making them feel competent and in control) builds trust, which in turn significantly boosts their intention to use the technology.
- Technology Specificity: The model explained a high degree of variance (69.3% for interaction mode and 66.0% for virtual image), but the weight of each predictor shifted. This proves that AV designers cannot use a "one-size-fits-all" psychological strategy for different vehicle features.
Key Results
- Preference for Function: Participants showed a significantly higher intention to use the functional Interaction Mode ($M=5.73$) compared to the Virtual Image ($M=5.26$).
- Trust Levels: Trust was higher for the functional alert system ($M=5.21$) than for the virtual assistant ($M=4.83$).
- Predictive Power:
- For the Interaction Mode, Autonomy, Relatedness, and Trust were the strongest predictors.
- For the Virtual Image, Autonomy and Trust were the primary drivers, while the direct effects of Competence and Relatedness were much weaker or insignificant.
- Mediation: Bootstrap analysis confirmed that automation trust partially mediated the effect of SDT needs on behavioral intention for both technologies.
Limitations and Future Work
- Age Bias: The sample was restricted to young and middle-aged adults (20–40 years old). Older drivers, who may have different trust thresholds or psychological needs regarding technology, were not represented.
- "First Contact" Bias: The study measured intentions based on video introductions. The authors acknowledge that psychological needs and trust are dynamic and may change after long-term, real-world use of an AV.
- Scope: The study focused on the interaction media rather than the intention to use the autonomous vehicle as a whole. Future work should explore if a positive interaction with a virtual assistant can "halo" into higher trust for the car's actual driving capabilities.
Summary updated