Dataset Information

Comparing supervised and unsupervised approaches to emotion categorization in the human brain, body, and subjective experience.

ABSTRACT: Machine learning methods provide powerful tools to map physical measurements to scientific categories. But are such methods suitable for discovering the ground truth about psychological categories? We use the science of emotion as a test case to explore this question. In studies of emotion, researchers use supervised classifiers, guided by emotion labels, to attempt to discover biomarkers in the brain or body for the corresponding emotion categories. This practice relies on the assumption that the labels refer to objective categories that can be discovered. Here, we critically examine this approach across three distinct datasets collected during emotional episodes-measuring the human brain, body, and subjective experience-and compare supervised classification solutions with those from unsupervised clustering in which no labels are assigned to the data. We conclude with a set of recommendations to guide researchers towards meaningful, data-driven discoveries in the science of emotion and beyond.

SUBMITTER: Azari B

PROVIDER: S-EPMC7679385 | biostudies-literature | 2020 Nov

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Comparing supervised and unsupervised approaches to emotion categorization in the human brain, body, and subjective experience.

Azari Bahar B Westlin Christiana C Satpute Ajay B AB Hutchinson J Benjamin JB Kragel Philip A PA Hoemann Katie K Khan Zulqarnain Z Wormwood Jolie B JB Quigley Karen S KS Erdogmus Deniz D Dy Jennifer J Brooks Dana H DH Barrett Lisa Feldman LF

Scientific reports 20201120 1

Machine learning methods provide powerful tools to map physical measurements to scientific categories. But are such methods suitable for discovering the ground truth about psychological categories? We use the science of emotion as a test case to explore this question. In studies of emotion, researchers use supervised classifiers, guided by emotion labels, to attempt to discover biomarkers in the brain or body for the corresponding emotion categories. This practice relies on the assumption that t ...[more]

PMID: 33219270

Similar Datasets

Project description:Sophisticated animal-borne sensor systems are increasingly providing novel insight into how animals behave and move. Despite their widespread use in ecology, the diversity and expanding quality and quantity of data they produce have created a need for robust analytical methods for biological interpretation. Machine learning tools are often used to meet this need. However, their relative effectiveness is not well known and, in the case of unsupervised tools, given that they do not use validation data, their accuracy can be difficult to assess. We evaluated the effectiveness of supervised (n = 6), semi-supervised (n = 1), and unsupervised (n = 2) approaches to analyzing accelerometry data collected from critically endangered California condors (Gymnogyps californianus). Unsupervised K-means and EM (expectation-maximization) clustering approaches performed poorly, with adequate classification accuracies of <0.8 but very low values for kappa statistics (range: -0.02 to 0.06). The semi-supervised nearest mean classifier was moderately effective at classification, with an overall classification accuracy of 0.61 but effective classification only of two of the four behavioral classes. Supervised random forest (RF) and k-nearest neighbor (kNN) machine learning models were most effective at classification across all behavior types, with overall accuracies >0.81. Kappa statistics were also highest for RF and kNN, in most cases substantially greater than for other modeling approaches. Unsupervised modeling, which is commonly used for the classification of a priori-defined behaviors in telemetry data, can provide useful information but likely is instead better suited to post hoc definition of generalized behavioral states. This work also shows the potential for substantial variation in classification accuracy among different machine learning approaches and among different metrics of accuracy. As such, when analyzing biotelemetry data, best practices appear to call for the evaluation of several machine learning techniques and several measures of accuracy for each dataset under consideration.

Project description:Seamlessly extracting emotional information from voices is crucial for efficient interpersonal communication. However, it remains unclear how the brain categorizes vocal expressions of emotion beyond the processing of their acoustic features. In our study, we developed a new approach combining electroencephalographic recordings (EEG) in humans with a frequency-tagging paradigm to 'tag' automatic neural responses to specific categories of emotion expressions. Participants were presented with a periodic stream of heterogeneous non-verbal emotional vocalizations belonging to five emotion categories: anger, disgust, fear, happiness and sadness at 2.5 Hz (stimuli length of 350 ms with a 50 ms silent gap between stimuli). Importantly, unknown to the participant, a specific emotion category appeared at a target presentation rate of 0.83 Hz that would elicit an additional response in the EEG spectrum only if the brain discriminates the target emotion category from other emotion categories and generalizes across heterogeneous exemplars of the target emotion category. Stimuli were matched across emotion categories for harmonicity-to-noise ratio, spectral center of gravity and pitch. Additionally, participants were presented with a scrambled version of the stimuli with identical spectral content and periodicity but disrupted intelligibility. Both types of sequences had comparable envelopes and early auditory peripheral processing computed via the simulation of the cochlear response. We observed that in addition to the responses at the general presentation frequency (2.5 Hz) in both intact and scrambled sequences, a greater peak in the EEG spectrum at the target emotion presentation rate (0.83 Hz) and its harmonics emerged in the intact sequence in comparison to the scrambled sequence. The greater response at the target frequency in the intact sequence, together with our stimuli matching procedure, suggest that the categorical brain response elicited by a specific emotion is at least partially independent from the low-level acoustic features of the sounds. Moreover, responses at the fearful and happy vocalizations presentation rates elicited different topographies and different temporal dynamics, suggesting that different discrete emotions are represented differently in the brain. Our paradigm revealed the brain's ability to automatically categorize non-verbal vocal emotion expressions objectively (at a predefined frequency of interest), behavior-free, rapidly (in few minutes of recording time) and robustly (with a high signal-to-noise ratio), making it a useful tool to study vocal emotion processing and auditory categorization in general and in populations where behavioral assessments are more challenging.

Dataset Information

Comparing supervised and unsupervised approaches to emotion categorization in the human brain, body, and subjective experience.

Publications

Comparing supervised and unsupervised approaches to emotion categorization in the human brain, body, and subjective experience.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets