Yutong Chen

Final Presentation

Final Project

How to Use AI to Analyze Information Collected for Understanding Complex Nutrition Data Visualizations

Key words: Data Visualizations, Qualitative data analysis, AI analysis, machine learning 

Project Description

Data visualization plays a pivotal role in how information is communicated and understood, significantly influencing user perceptions and the effectiveness of data communication. These visualizations are constructed using various elements like visual encoding, chart types, clarity, data integrity, and narrative, each of which plays a critical role in shaping how information is received.

Understanding the perception of data visualizations is crucial as it influences how effectively information is communicated and comprehended [1]. Visual encoding utilizes aspects such as color, shape, position, and size, which can either simplify complex data for better accessibility or increase cognitive load, making data overwhelming. Different types of charts like bar graphs and scatter plots are selected based on their suitability for the data structure or the relationships they need to highlight, which directly influences user engagement and comprehension levels. Clarity is paramount and is achieved through simplicity in design, clear labeling, and maintaining consistency across the visualization, ensuring that the information is straightforward and easy to understand. Maintaining data integrity and providing accurate representations are essential for building user trust and credibility. A strong narrative that includes relevant contextual information and effective storytelling can enhance the emotional response and retention of information. For instance, using a consistent color scheme can reduce cognitive load and a well-designed scatter plot can expose correlations that might be missed in simpler representations. Ethical considerations are also crucial. Data visualizations should be created with transparency, accuracy, and inclusivity to avoid misleading representations and ensure they are accessible to all users. These considerations include addressing how visualizations are perceived when generated by AI versus those created by humans [2, 3].

In the field of nutrition, data visualization is extensively utilized to communicate research findings and educate the public, using tools like MyPlate, informational sheets, and various educational materials. With advancements in machine learning (ML) and natural language processing (NLP), it is now possible to create more tailored and comprehensible graphs and infographics suited to different audiences. Despite much research focusing on the development of AI-generated data visualizations, there is limited research on how these visualizations are perceived compared to those created by humans. This study aims (1) to compare the comprehension, perception, and effectiveness of AI-generated versus human-created visualizations, (2) to assess how different levels of descriptions in human-created graphs affect understanding, and (3) to analyze responses collected regarding the barriers to comprehending nutrition data visualizations using NLPs.

This study seeks to fill the research gap by comparing the perception of AI-generated versus human-created visualizations in nutrition data. By analyzing responses collected via platforms like Qualtrics, this research will assess comprehension and perception differences across various types of visualizations. The study will involve students and postdoctoral fellows from Tufts University, exposing them to different levels of descriptive detail in human-created graphs and AI-generated graphs. A mixed-methods approach, combining quantitative surveys with qualitative focus groups, will be used to gather comprehensive feedback on the visualizations. I will use NLPs to analyze the survey and focus group responses. Anticipated results are expected to provide insights into how different visualization types are perceived in terms of clarity, evidence, emphasis, and efficiency. This could reveal significant differences in comprehension levels, highlighting the strengths and weaknesses of each visualization type and influencing the design of future educational materials in nutrition.

By addressing these aspects, the study not only aims to enhance how nutritional information is communicated to the public but also contributes to the broader body of knowledge regarding the application of AI in data visualization. This research will ultimately aid in improving educational outcomes and public health, making nutritional data more effective and accessible.

AI Ethics Statement

Ensuring AI systems used for creating visualizations are transparent and accountable is crucial for the integrity of the research. Participants should be informed about how these visualizations are generated, and researchers must ensure their accuracy and integrity. Transparent processes help build trust and allow for proper oversight. AI-generated visualizations should also be fair and equitable, avoiding biases that could misrepresent data or mislead interpretations. This involves careful consideration of the training data and algorithms used to ensure they do not aggravate existing inequalities or introduce new ones. Visualizations with AI-generated descriptions should also be culturally sensitive, accurately representing data in a way that is respectful and relevant to diverse audiences, considering cultural differences in data interpretation and presentation styles. Transparency and accountability are the most critical aspects because they underpin the trustworthiness of AI-generated visualizations; without clear processes and accountability, users may doubt the validity of the information presented. Fairness and equity are equally crucial, as biased visualizations can lead to misinformed decisions and perpetuate health disparities [4]. To address these ethical issues, a comprehensive policy framework involving transparency to participants should be implemented, detailing how the data visualization is developed [5]. Addressing these ethical issues will lead to more reliable, fair, and culturally sensitive nutritional data visualizations, enhancing public trust and improving educational outcomes.

Knowledge Graph

Ontology:

Conceptual/Causal Diagrams:

References:

1.         Pineo, D. and C. Ware, Data visualization optimization via computational modeling of perception. IEEE transactions on visualization and computer graphics, 2011. 18(2): p. 309-320.

2.         Etemadpour, R., et al. Role of human perception in cluster-based visual analysis of multidimensional data projections. in 2014 International Conference on Information Visualization Theory and Applications (IVAPP). 2014. IEEE.

3.         Few, S., 35. Data visualization for human perception. The Encyclopedia of Human-Computer Interaction, 2013.

4.         Israel, M.J. and A. Amer, Rethinking data infrastructure and its ethical implications in the face of automated digital content generation. AI and Ethics, 2023. 3(2): p. 427-439.

5.         Bendel, O., Image synthesis from an ethical perspective. AI & SOCIETY, 2023: p. 1-10.