DiG: An interface for goal-oriented decision making

DiG was an experimental interface for decision-making. While there are many web sites and interaction techniques for filtering, sorting, visualizing and summarizing data, most of them require users to interact with low-level features. DiG, on the other hand, attempts to determine a user's goals, connects those goals to low-level features behind-the-scenes, and then shows the user a set of decisions that include the features that are important to meet their goals.

That's the idea anyway. When rubber hits road the result is always more mundane (and more difficult and interesting) than rhetoric. This project is no different. The implementation focused on helping users select a camera.

This was an unusual hybrid interface combining a Java applet and an HTML5 widget developed by Aditi Muralidharan.

1

To get started users first answer a set of very simple questions that focused on how they planned to use the product. Clicking on a question opened a multiple-choice selection list of uses that are derived automatically or semi-automatically from mining user review data. The uses themselves could be clustered into questions, but we assume that a person familiar with the domain would do a much better job grouping them.

2

The selection of uses indirectly selected low-level aspects that had been associated with the uses a priori. After answering the high-level questions users got to see which camera aspects best map to their stated goals. Aspects were also weighted, so for example to meet some goals (hiking) the aspect "durability" would be heavily weighted whereas for others (formal dinners) it would not matter at all.

The aspects themselves were derived from user review data as well as product sheet information (e.g., "resolution").

At this point users could refine their search by adjusting aspect weights or swapping in aspects the system had not selected at all.

Francine Chen did the bulk of the analysis deriving aspects from reviews and connecting them to uses.

3

All of the low-level data was available for each camera, including reviews and technical specifications. The interface also allowed them to add cameras to a save-list. An Android client allowed them to copy camera data to their mobile. This feature was added with the idea that the interface might be positioned on a large display at the front of the store. Once a user selected her cameras she could move them onto her phone to go find them and try them out.

4

DiG isn't the only goal-oriented decision system. Nutmeg is an online investment manager that takes a very similar approach, asking users some simple questions about their financial goals up front and then determining a portfolio that satisfies those goals. The tool goes a step further than DiG and automatically adjusts the "weight" of underlying "aspects" (balance of funds) as conditions change.

Given the increasing complexity of choices in day-to-day life, Nutmeg and companies like it seem like the future of decision management.