Case Study: Looxid Emotion AI System in User Testing Solution

Discovering Consumer’s Unspoken Emotion within Immersive Enterprise VR Experience

In our earlier posts, we have introduced a number of different methodologies that companies can use to better understand consumers.
Companies identify consumer preferences at every stage of the value chain, from product development to service, and reflect them in strategic decision making. When using conventional consumer research methods, however, responses may be biased or untruthful for several reasons. People in the industry and the academic world questioned the reliability was questioned. And hence, alternatives were proposed, such as neuromarketing and emotion recognition technologies.

The core value of Looxid Labs solution is in discovering consumer’s unspoken emotions through biological responses measured during a VR experience. In this experience with a VR headset, the face of the user touches the device directly. So it has become easier to naturally monitor user’s biological signal that includes a lot of useful information. When the user wears our VR headset, the embedded sensors in LooxidVR seamlessly measure brain activity and eye movement. Based on the brain and eye data collected, an emotion AI learns about user’s emotional status and builds a personalized model. Then this model is applied to a content which a client requests for analysis. We can use the AI system to evaluate client prototypes, interior designs, marketing videos, and more.

The following cases are some usage examples of Looxid emotion AI system. The first one is an experiment on people(30 participants)’s response to an anti-smoking advertisement. The heat map on the video displays the gaze-tracking analysis which points out where participants primarily looked. The first and the second graph represent arousal and valence analysis respectively.

Participants’ arousal level was the highest when watching a painful face of the woman. This declined gradually, but as her painful face reappeared in the video, it went up again. Overall, the level of valence decreased throughout the experiment. From the results, we can infer that participants’ negative feeling towards smoking increase while watching the video. This kind of information will be useful when trying to maximize the effectiveness of an advertisement.

Next case is a pilot project on car design evaluation. A total of 10 participants were asked to observe the exterior of an SUV vehicle in a VR environment. Then, a written questionnaire(7-point Likert scale) on design preference and a qualitative interview followed. The heat map in the image below represents participants’ average preference level by location, which is a combined topography of the user’s eye(gaze, pupil, saccade) and brainwave data.

We divided the car into a front, side, and rear area in the analysis of physiological signals. The blue bar shows the average preference of the participants based on their survey answers, and the gray bar is the result of AI analysis. As you can see, the two show a similar tendency.

But if we look at the individual participants, there is a slight difference between the two methods of analysis, questionnaire and AI. For instance, participant #7 answered that he preferred the front part the most in the survey, but our AI did not agree. Participant #2 responded that he gave high points to the overall vehicle design in the survey, but not in our analysis. For reference, participant #7 commented in the post-VR interview, “I wish the front grille looked more durable,” and participant # 2 said, “I didn’t feel up for buying this SUV since its design was too common.”

The cases mentioned above are some ways of applying our solution. We are developing an emotion AI solution that provides our clients with quantitative insights into user’s emotion by measuring human physiological data during a VR experience. Looxid Labs’ technology will enhance business organization’s confidence in making strategic decisions. In our next post, we will look deeper into the reason behind the difference between AI analysis and survey results, and also some examples of insight you can gain from emotion AI analysis.