The Legilimens! A scientific spell using neuroscience to read your mind: real-time emotion estimation application and its implications in the cosmetic industry
Podium 30
Presented by: Gusang Kwon
Introduction
We have witnessed loads of changes in our life after the COVID-19 pandemic, including in the beauty and cosmetic industry. Most people progressively started focusing on their innermost feelings and pursuing mental wellbeing. The increase in fragrance product sales confirms this trend. Many consumers were increasingly seeking and buying fragrance products, including perfumes, during the pandemic period than before. Such products, like lipstick, became a small luxury to young consumers. We use fragrance for relaxing, refreshing, and sometimes when we are depressed. In the era of restricted traveling, fragrance can act as a mental getaway to escape from reality, unlike before. Fragrance can bring in specific moments of time or space as a sense of smell is strongly related to memory. Based on the importance of this aspect, research on consumer behavior towards fragrance has also developed.
The sense of smell has a strong impact on our physiology and emotions. Understanding the exposure to fragrances via the olfactory system in the brain has been well documented in the literature [1-3]. Thanks to the development of neuroscientific techniques, we can nowadays estimate various emotions with better accuracy [4]. However, identifying specific emotions such as anger or joy can be difficult as they do not seem to have a universal neural fingerprint across people [5]. Therefore, we instead focused on the dimensional model of emotions, particularly on the idea of mapping the emotional state into a two-dimensional space of valence and arousal. Valence (positive or negative) and arousal (aroused or relaxed) are well-defined factors in describing emotional experience [6]. Previous studies have shown that it is feasible to detect the states of arousal and valence from EEG signals [7,8].
In this study, we introduced a real-time emotion estimation application and its use in cases using fragrance. In a way, we can call it a ‘scientific spell,’ which is pretty similar to the ‘Legillimens’ spell in Harry Potter, the practice of using magic to enter into another person’s mind.
Method
We used a wearable EEG headset (MOOD8, imec, Eindhoven, The Netherlands), which is a wireless and multimodal research-grade device in combination with soft dry polymer-coated Ag/AgCl electrodes (SoftPulse™, Datwyler, Altdorf, Switzerland) [9]. One hundred healthy adults participated in this study (age range: 20-52; F = 82, M = 18, IRB approved: 2021-1CR-N21P). Subjective assessments of valence, arousal, and preference were captured via Likert scales, using the Psychopy GUI. Six fragrances were administered and each fragrance was given three times. All the fragrances and their iterations were randomized. Each trial lasted 15 seconds and the participants were asked to fill in the scales right after each fragrance. We used conventional EEG analysis methods and machine-learning algorithms to find applicable features for the application.
Results
We developed a windows application to work with imec's MOOD8 headsets. It can read and record EEG data from the headset, plot it in real-time, run the algorithm pipeline on the data and finally show the participant emotions in real-time (arousal, valence, and preference). This app brings a smooth usability flow where the main tasks are divided into a stepwise approach to not overwhelm the user by displaying all the functionality on one screen.
Discussion and Conclusion
In this project, we presented a headset device for acquiring EEG data and a portable application for visualizing the data and insights created by the algorithms wrapped inside the app. An essential part of the algorithm is a machine learning model pre-trained using collected 100 subjects’ data and self-reported emotional responses.
We can use this application in various steps, e.g., product development, consumer survey, and marketing. Further, this can solve the purpose of increased need for a tailored solution. This tool can help understand consumer behavior and mind in a novel way, thereby contributing and providing multiple implications in the cosmetic industry, as well as neuromarketing and consumer neuroscience fields.
References
[1] Ilmberger, Josef, et al. Chemical Senses 26.3 (2001): 239-245.
[2] DOI: 10.1016/S0944-7113(98)80041-X.
[3] DOI: 10.1016/j.eujim.2015.08.006.
[4] N. S. Suhaimi et al., Computat. Intell. Neurosci., 2020 (2020).
[5] DOI: 10.1017/S0140525X11000446.
[6] DOI: 10.1515/REVNEURO.2004.15.4.241.
[7] DOI: 10.1007/978-3-642-38803-3_6.
[8] DOI: 10.1109/ACII.2015.7344552.
[9] https://www.imec-int.com/en/articles/amorepacific-uses-imecs-eeg-headset-neuromarketing-research, (2020)
We have witnessed loads of changes in our life after the COVID-19 pandemic, including in the beauty and cosmetic industry. Most people progressively started focusing on their innermost feelings and pursuing mental wellbeing. The increase in fragrance product sales confirms this trend. Many consumers were increasingly seeking and buying fragrance products, including perfumes, during the pandemic period than before. Such products, like lipstick, became a small luxury to young consumers. We use fragrance for relaxing, refreshing, and sometimes when we are depressed. In the era of restricted traveling, fragrance can act as a mental getaway to escape from reality, unlike before. Fragrance can bring in specific moments of time or space as a sense of smell is strongly related to memory. Based on the importance of this aspect, research on consumer behavior towards fragrance has also developed.
The sense of smell has a strong impact on our physiology and emotions. Understanding the exposure to fragrances via the olfactory system in the brain has been well documented in the literature [1-3]. Thanks to the development of neuroscientific techniques, we can nowadays estimate various emotions with better accuracy [4]. However, identifying specific emotions such as anger or joy can be difficult as they do not seem to have a universal neural fingerprint across people [5]. Therefore, we instead focused on the dimensional model of emotions, particularly on the idea of mapping the emotional state into a two-dimensional space of valence and arousal. Valence (positive or negative) and arousal (aroused or relaxed) are well-defined factors in describing emotional experience [6]. Previous studies have shown that it is feasible to detect the states of arousal and valence from EEG signals [7,8].
In this study, we introduced a real-time emotion estimation application and its use in cases using fragrance. In a way, we can call it a ‘scientific spell,’ which is pretty similar to the ‘Legillimens’ spell in Harry Potter, the practice of using magic to enter into another person’s mind.
Method
We used a wearable EEG headset (MOOD8, imec, Eindhoven, The Netherlands), which is a wireless and multimodal research-grade device in combination with soft dry polymer-coated Ag/AgCl electrodes (SoftPulse™, Datwyler, Altdorf, Switzerland) [9]. One hundred healthy adults participated in this study (age range: 20-52; F = 82, M = 18, IRB approved: 2021-1CR-N21P). Subjective assessments of valence, arousal, and preference were captured via Likert scales, using the Psychopy GUI. Six fragrances were administered and each fragrance was given three times. All the fragrances and their iterations were randomized. Each trial lasted 15 seconds and the participants were asked to fill in the scales right after each fragrance. We used conventional EEG analysis methods and machine-learning algorithms to find applicable features for the application.
Results
We developed a windows application to work with imec's MOOD8 headsets. It can read and record EEG data from the headset, plot it in real-time, run the algorithm pipeline on the data and finally show the participant emotions in real-time (arousal, valence, and preference). This app brings a smooth usability flow where the main tasks are divided into a stepwise approach to not overwhelm the user by displaying all the functionality on one screen.
Discussion and Conclusion
In this project, we presented a headset device for acquiring EEG data and a portable application for visualizing the data and insights created by the algorithms wrapped inside the app. An essential part of the algorithm is a machine learning model pre-trained using collected 100 subjects’ data and self-reported emotional responses.
We can use this application in various steps, e.g., product development, consumer survey, and marketing. Further, this can solve the purpose of increased need for a tailored solution. This tool can help understand consumer behavior and mind in a novel way, thereby contributing and providing multiple implications in the cosmetic industry, as well as neuromarketing and consumer neuroscience fields.
References
[1] Ilmberger, Josef, et al. Chemical Senses 26.3 (2001): 239-245.
[2] DOI: 10.1016/S0944-7113(98)80041-X.
[3] DOI: 10.1016/j.eujim.2015.08.006.
[4] N. S. Suhaimi et al., Computat. Intell. Neurosci., 2020 (2020).
[5] DOI: 10.1017/S0140525X11000446.
[6] DOI: 10.1515/REVNEURO.2004.15.4.241.
[7] DOI: 10.1007/978-3-642-38803-3_6.
[8] DOI: 10.1109/ACII.2015.7344552.
[9] https://www.imec-int.com/en/articles/amorepacific-uses-imecs-eeg-headset-neuromarketing-research, (2020)