Automated Disclosure Literature Review
Context
Discover recently launched Disclosure Playback, an automated system that reads mandatory disclosures to customers. This helps reduce compliance risk by ensuring disclosures are delivered accurately and consistently, rather than relying solely on agents to read them manually.
While this system addresses operational and compliance needs, the team also wanted to protect the strong customer experience Discover is known for.
The objective of this project was twofold: (1) to identify what is already known about customer perceptions and preferences related to automated disclosure systems, and (2) to uncover gaps in knowledge that would inform future research. To do this, I conducted a comprehensive literature review, providing the team with evidence-based insights to guide decisions about system design and implementation.
Tools: n/a
Method Used: Literature Review
Participants: 4 Participants who use Assistive Technologies
Timeline: 2 Weeks
Why a Literature Review
I began with a literature review to quickly surface existing evidence and guide early decisions without the time and cost of new research. Drawing on my background in experimental psychology, I reviewed studies from journals like American Speech, Human Factors and Ergonomics, and CHI. This approach provided evidence-based insights for system design and identified key gaps in our experience.
Key Takeaways
👂🏼 Participants were more satisfied with automated systems that used natural, human-like speech.
Key Finding: Aspects of human speech such as adjustments in pitch and intensity correspond with social likability, cooperation and affinity in systems (International Journal of Human-Computer Studies).
Design Consideration: Ensure that automated disclosures reflect aspects of human such, such as natural inflection and fluidity.
🤖 Preferences for automated disclosures vary, making a “one-size-fits-all” design challenging
Key Finding: Study of in-car interaction with voice assistances found that users trusted and liked voice systems more when the system’s personality matched their own. (Proceedings of CHI Conference on Human Factors in Computing Systems).
Design Consideration: Prioritizing simplicity and clarity in disclosure design may be more effective than attempting to accommodate for user preferences.
🗣 While consumers say they prefer female voices in system messages, voice type doesn’t affect disclosure rates in sensitive contexts
Key Finding: Some research has found that female voices are perceived as more attractive (American Speech), however, medical studies found that disclosure rates were not affected by gender in a medical setting, and patients equally disclosed sensitive medical information to male and female voices (Interacting with Computers).
Design Consideration: Using predominantly female-voices in systems may push harmful gender stereotypes. It’s important not to prioritize user preferences too much, as they don’t significantly impact disclosure rates.
🔐 Consumers find privacy disclosures hard to understand but view their complexity as a sign of stronger security
Key Finding: Consumers often perceive traditional paragraph-style privacy policies as more secure, yet struggle to understand them. A survey found that 17% avoid reading these notices due to their length, legal jargon and complexity (Considering Consumer Privacy: A Resources for Policymakers and Practitioners).
Design Consideration: To ensure they understand the information, consider asking customers if they have any questions or need clarification at the end of the disclosure.