Blog 05: Mental Health Perspective of Affective Computing

Affect as information is the process where interactive systems gather information from individuals and further asses to provide useful suggestions. Imagine talking about this 20 years in the past. As technology has emerged, we are discovering new ways to ease our lives. Today we already have personal information collected to analyze and provide insights for our physical data. With wearables, users have detailed health data available on their fingertips. Affect as information allows us to obtain similar insights for our emotions. This may sound like a very straight forward approach but implementation tells us a different story. Emotions are something which could be very tricky to evaluate. Here, I would like to share an example from my professor Dr. Andrea Kleinsmith. She mentioned one instance from one of her studies in which they were trying to understand the facial emotions of users when they were angry. The first thought that came to the majority of the students' minds was frowning eyebrows. Surprisingly, the results indicated that majority participants smiled when they were frustrated. This was something I was not expecting but at the same time I wasn’t surprised to hear as well. Haven’t we all smiled in anger at least once? In my opinion, the fact that an individual does not react  to similar situations in the same way every time is one of the biggest challenges to implementing affect as info.

To further understand the area of affect as information I studied two research papers this week. First study was named “Toward an Affect Sensitive AutoTutor”. This study made use of an affective loop to develop an affect sensitive tutoring system with the ability to recognize, assess, and react towards a learner’s affective state. The system was named affective tutoring and was aimed to simulate real tutors with natural language processing. The affective tutor taught students newtonian physics and computer literacy topics by presenting challenging problems  from a curriculum script and engaging students in a mixed-initiative dialogue to construct an answer. The literature for affect as information prior to this study was focused mainly on six major factors like fear, anger, happiness, sadness, disgust, and surprise. With this study the authors made an attempt to widen the scope for research by exploring additional factors. The results of this study claimed that engaging users with activities which increase interest, enhance simulation, offer choices and challenges may result in improving affect as information.

Moving on to the second study, “AffectAura: An Intelligent System for Emotional Memory” was a study based on the concept of lifelogging that allowed the users to reflect on their emotional states over long periods of time. This was done by collecting audio, visual, physiological and contextual data for predicting users affective state and providing users which reflected their emotions. The results of this study contributed towards a memory aid to record events automatically and help the user recall events upon reflection. This was something which laid a foundation for future work for assisting memory aid to help users to hold on to their life learnings which are often forgotten in a day or two.

This week's learnings helped me realize how information can help us assess our emotions which can contribute towards assessing our emotions. Today, affect as information is seeing new trends with AI able to assess our emotions better than ourselves resulting in providing us with appropriate suggestions. Talking about the current AI and ML capabilities, I have had personal experiences with Spotify's recommendation system to be able to suggest a song which I myself couldn’t think of being appropriate for my mood. With this capability of AI and ML algorithms merging with affect, I look forward to seeing what affect as information holds for us in the future. I believe that this would result in solving a wide range of problems which mainly include advancements in the field of mental health and betterment of society ONLY. While I end this blog here, my only concern would be the tech giants misuse these technologies to make more money and find ways to sell us things which we do not need.  

References:

[1]AffectAura: An Intelligent System for Emotional Memory

[2]Toward an AffectSensitive AutoTutor