Martin Wattenberg: ML Visualization and Interpretability

Balancing Info

Balancing complexity and simplicity in visual design is crucial for effective communication. Martin Wattenberg emphasizes the importance of understanding the audience when deciding how much information to include in a visualization. For newcomers, simplicity is key, while experts may benefit from more detailed data 1. Progressive disclosure is a technique that allows users to access more information as needed, catering to both novices and experts 2. Wattenberg notes, "The biggest mistake is people try to make things too simple more than anything else."

The biggest mistake is people try to make things too simple more than anything else.

--- Martin Wattenberg

This approach ensures that visualizations are both informative and accessible, adapting to the user's level of expertise.

   

Multi-Scale Models

Multi-scale models in information graphics enhance visualization by organizing data hierarchically. Martin Wattenberg explains that effective design guides the viewer's eye from large structures to finer details, making complex information more digestible 3. This approach requires understanding the user's mental model and leveraging the visual system's capabilities. However, Wattenberg acknowledges the limitations of quantitative methods, emphasizing the value of qualitative insights in design 4. He states, "Qualitative understanding, just talking to someone and watching them use a visualization for five minutes is probably going to be more informative than plugging the image into a fancy mathematical model."

Qualitative understanding, just talking to someone and watching them use a visualization for five minutes is probably going to be more informative than plugging the image into a fancy mathematical model.

--- Martin Wattenberg

This highlights the need for a balanced approach that combines both quantitative and qualitative methods in visualization design.