1. First, the computer presents several running shoe design proposals to the user.
2. Next, the user evaluates the proposed designs based on their own sensibilities and preferences, with reactions such as “I like this design!” or “This one isn’t quite right…”
3. Then, the user’s feedback is sent back to the computer, which generates new design proposals based on the evaluations and presents them again to the user.

In recent years, product development fields have increasingly sought to create items that align with user sensibilities, and the application of IEC is gaining attention. However, a major challenge in IEC is the high evaluation burden placed on users. To reduce this burden, various approaches have been explored, including improvements to algorithms and evaluation interfaces. Our research focuses on reducing user evaluation burden through enhancements to the evaluation interface.

In conventional IEC, the typical evaluation interface requires users to assign a rating—such as on a 5-point or 10-point scale—to about 10 to 20 presented designs. However, this can cause users to hesitate in assigning scores, leading to a significant evaluation burden.

We therefore proposed an evaluation interface that avoids the cumbersome task of assigning numerical ratings. Instead, users simply select only the designs they prefer from the ones presented. To date, we have introduced several such interfaces, including the “Tournament-Based Evaluation Method,” where users evaluate designs in a bracket-style format, and “Interactive Tabu Search,” where users select only one preferred design from multiple options. In addition, we are also working on developing evaluation interfaces based on eye gaze and conversational input.

Evaluation via Tournament Format

In the tournament-based evaluation method, the user repeatedly compares two presented designs and selects the one they prefer. This approach is especially effective in reducing the evaluation burden when assessing time-based data such as music or video, as only two items are compared at a time.
There are two types of tournament-based evaluation methods: the "Standard Tournament Format," where the user selects their preferred design from the two presented, and the "Graded Tournament Format," where the user evaluates the two designs with nuanced preferences such as “strongly prefer the left design” or “slightly prefer the right design.”

Select Only One Preferred Design

In Interactive Tabu Search, the user selects only one preferred design from a set of presented options. While it is less suitable for evaluating time-based media such as music or video—due to the need to compare multiple items at once—it simplifies the evaluation process for static images, effectively reducing the user’s evaluation burden.

Evaluation Experiments

Up to now, we have conducted both numerical simulations—using evaluation agents generated on a computer in place of real users—and user-based evaluation experiments. The numerical simulations were used to assess the fundamental performance of the proposed methods, while the user experiments focused on evaluating ease of use from the users’ perspective.
These results indicate that both the Tournament-Based Evaluation Method and Interactive Tabu Search are effective in reducing the evaluation burden on users in IEC.