Developing Design Guidelines and Frameworks for Human-Centered Conversational Voice AI
Graduate Research
Project Summary
- Developed and validated 14 conversational AI design guidelines synthesized from 336 published recommendations, creating a practical framework used in workshops and professional training for UX designers
- Planned and executed interviews, surveys, usability evaluations, heuristic evaluations, cognitive walkthroughs, and Delphi studies across multiple research programs
- Disseminated developed guidelines across several public courses and workshops with both academic and industry designers
- Published and presented findings and recommendations at leading international HCI venues including CHI, CSCW, MobileHCI, CUI, and IEEE conferences
- Collaborated with researchers, designers, and industry professionals to bridge research insights with real world design practice through courses and workshops
❓️
Research Problems:
- What are the practices, workflows, and barriers current industry VUI designers experience?
- How do we take the myriad of VUI heuristics proposed in academic literature and synthesize them into a cohesive set of industry-friendly design principles that industry designers can adopt into their workflows
💡
Research Goal
To identify the major barriers and needs of industry VUI designers, and using that knowledge, take the academic research that has been conducted in VUI design, and synthesize them into industry-friendly heuristics that can support industry designers in creating usable VUIs
14 Synthesized Guidelines for Designing Conversational Voice User Interfaces (from Murad et al., 2023)
| G# | Guideline | Description |
|---|---|---|
| V1. | Design conversation based on the task domain of the voice interface. | Conversations should be tailored to the task domain that the voice interface has been made for, and it should contain an extensive coverage of knowledge in that task domain. |
| V2. | Personalize the user experience to each user based on context. | A VUI should use context and background knowledge from the user and the task domain in order to personalize the user experience for every user, and to allow users to personalize their own user experience. |
| V3. | Be transparent about system intelligence, capabilities, and constraints. | A VUI should be transparent about what its capabilities and its constraints are. It should clearly state what it can, and cannot do, to properly manage a user’s expectations about the system’s intelligence and abilities. |
| V4. | Break tasks up to allow for less cognitive load and better recall. | A VUI should break up tasks into more manageable sub-tasks, to require less cognitive load, and allow users to recall information and functionality more easily. Menus should be short, and dialogue prompts should be minimal in nature. |
| V5. | Design clear and informative feedback. | VUIs should design feedback to be clear and easy to understand, while also providing all the necessary information required from said feed0back – such as when the system is processing information, when errors occurred, etc. Feedback should be provided quickly and efficiently to the user. |
| V6. | Design appropriately for specific user and application types. | A VUI should be designed based on the user that will be interacting with it, and the type of application it will be used in. |
| V7. | Protect the users’ privacy and security throughout conversational interaction. | Interactions and information shared with a VUI should always be kept private, and security measures should be built throughout a conversational interaction (authorization, authentication, etc.). |
| V8. | Design conversational interaction that maps to real-world conversational norms and dialogue patterns. | A VUI should use well-known and familiar conversational norms and dialogue patterns when interacting with a user and allow users to use natural speech and language to interact with it. A VUI should work to match realistic mental models of conversation, where appropriate (depending on the task domain). |
| V9. | Make conversation consistent and efficient across the user interaction experience. | Conversations should be designed to be consistent across the entire interface. Prompts should use the same type of formulation and language across the interface, and similar commands should use similar language to invoke them. The VUI should also be efficient and minimalist in conveying information to the user. |
| V10. | Design a persona that evokes positive emotions with users. | VUIs should provide a pleasant persona that provides a positive experience for users. The persona should use things like encouragement, engagement, and humor to evoke positive emotions with users, and help users feel more comfortable with being personal with the VUI. |
| V11. | Help prevent, recognize, and correct errors. | A VUI should work to prevent the user from encountering any errors, by providing appropriate help and documentation, and guiding the interaction to assist users in providing the correct input to be recognized. If an error occurs, the VUI should communicate the error quickly and efficiently, allowing the user to recognize what the error is, and allow the user mechanisms to correct errors. |
| V12. | Create diverse and accessible speech interaction. | A VUI should be designed to be accessible to many types of users (those with hearing/mobility impairments, for example), and be usable by a diverse range of users. |
| V13. | Provide open-ended prompts to establish common ground, and to request or clarify information. | Open-ended prompts should be used to establish a common ground with the user, and to obtain background knowledge of the user. Open-ended prompts should also be used to request input from users, with more restrictive prompts being later used to clarify information. |
| V14. | 4.14 Provide an interactive and intuitive user experience with a large amount of user control. | A VUI should be interactive and provide the user with as much control as possible over the user experience and its functions. A VUI should provide control over navigating through the interface as well. |
Resulting Publications:
- Christine Murad, Cosmin Munteanu, Leigh Clark, and Benjamin R. Cowan. 2018. Design guidelines for hands-free speech interaction. In Proceedings of the 20th International Conference on Human-Computer Interaction with Mobile Devices and Services Adjunct (MobileHCI '18). Association for Computing Machinery, New York, NY, USA, 269–276. https://doi.org/10.1145/3236112.3236149
- Christine Murad and Cosmin Munteanu. 2019. "I don't know what you're talking about, HALexa": the case for voice user interface guidelines. In Proceedings of the 1st International Conference on Conversational User Interfaces (CUI '19). Association for Computing Machinery, New York, NY, USA, Article 9, 1–3. https://doi.org/10.1145/3342775.3342795
- Christine Murad, Heloisa Candello, and Cosmin Munteanu. 2023. What’s The Talk on VUI Guidelines? A Meta-Analysis of Guidelines for Voice User Interface Design. In Proceedings of the 5th International Conference on Conversational User Interfaces (CUI '23). Association for Computing Machinery, New York, NY, USA, Article 19, 1–16. https://doi.org/10.1145/3571884.3597129
- Christine Murad. "Supporting Designers in Voice User Interface Design." PhD diss., University of Toronto (Canada), 2024.
- Irene Lopatovska and Christine Murad. 2026. Designed for Interfaces, Not Conversations: The Limits of Usability Heuristics in GenAI Chatbot Assessment. In Proceedings of the 8th ACM Conference on Conversational User Interfaces (CUI '26). Association for Computing Machinery, New York, NY, USA, Article 78, 1–6. https://doi.org/10.1145/3816046.3816270