Foody Talk: Exploring Opportunities for Conversational Food Journaling

Research Background and Problem

• What issues or challenges did the authors identify?
  Digital food journaling tools can help people reflect on and improve their eating habits. However, these tools are often perceived as burdensome, leading users to discontinue use after a short period and fail to fully benefit from them. Although some studies have attempted to reduce the complexity of logging, such as using voice input or photos, these methods still have limitations, such as data accuracy, input burden, and the tendency of popular technologies to promote packaged food consumption over fresh food.

• Why is this issue important?
  Food journaling has the potential to support health goals, such as improving eating habits, weight management, collaboration with nutritionists, diabetes management, and identifying and addressing food allergies or intolerances. However, the low usage rates and poor user experience of current tools hinder the widespread adoption and long-term adherence to food journaling.

• Research motivation and related work:
  Conversational User Interfaces (CUIs) are considered a promising solution, as they can support food journaling in a natural and interactive way. However, there is still insufficient understanding of how users wish to use CUIs' conversational features for logging and reflection. This paper aims to fill this gap and explore how CUIs can personalize and enhance the food journaling experience to better support users' goals.

Solution

• What methods or solutions did the authors propose?
  The authors proposed designing a method for food journaling using CUIs, which can learn users' goals, terminology, and eating habits, and optimize interactive dialogues to support logging and reflection. The study explored users' preferences for CUI interactions through 33 co-design sessions with 18 participants.

• What are the innovative aspects of this solution?

  1. Personalized learning: CUIs need to understand users' individual goals and terminology, such as food categories and meal times.
  2. Depth and interaction: CUIs should use iterative dialogues to explore users' food descriptions and health reflections, encouraging detailed logging.
  3. Contextual adaptability: CUIs should adjust interaction depth and feedback based on contextual constraints, such as time pressure or social settings.
  4. Emotional support: CUIs should provide encouraging and non-judgmental feedback during reflection to enhance the user experience.

• What are the implementation steps and key technologies used?
  Key steps include:

  1. Conducting participatory design activities using tools like Miro boards, exploring "ideal" dialogues through scenario cards and dialogue cards.
  2. Thematic analysis and resonance design to identify user expectations for CUIs in supporting data collection and reflection.
  3. Integrating machine learning techniques and large language models (LLMs) to help CUIs learn users' dietary terminology and habits.

Research Findings

• What specific outcomes were achieved?

  1. Data collection: Participants expected CUIs to construct goal-related logs from conversations, supplement details (e.g., ingredients, portions) through proactive questioning, and infer goals. Additionally, CUIs should support quick logging scenarios and adapt to contextual constraints like time pressure.
  2. Data reflection: CUIs can provide instant feedback (e.g., progress toward nutritional goals), support customized data queries (e.g., daily vegetable intake), and assist in exploring food-related issues (e.g., triggers for allergies or gastrointestinal discomfort).
  3. Emotional support and accountability: CUIs can offer positive, encouraging language and maintain a sense of user control in both structured and unstructured feedback.

• What advantages does it have compared to existing solutions?
  CUIs are more natural and interactive than traditional food journaling tools, promoting deeper user engagement. Additionally, compared to current machine learning models, CUIs demonstrate greater potential for personalization and emotional support by learning individual terminology and goals.

• What were the experimental or evaluation results?
  The CUI dialogues designed by participants revealed a preference for multi-turn interactions and trust in CUIs' ability to learn their everyday terminology and habits. Participants also believed that CUIs could promote healthier behaviors, such as regular eating patterns and addressing food-related discomfort.

• Limitations and future directions:
  Limitations:

  • CUIs may be inefficient and less competitive compared to traditional visual interfaces, such as statistical charts, when navigating large, long-term datasets.
  • The feasibility of implementing the proposed conversational models with current technology remains to be validated.

  Future directions:

  • Explore how CUIs can collaborate with families, teams, or nutritionists in social contexts (e.g., shared food journaling).
  • Develop hybrid models that combine CUIs with visualization interfaces to optimize long-term data navigation.
  • Consider CUI designs that accommodate diverse cultural contexts to meet global variations in dietary habits and expectations.