In the realm of artificial intelligence, deep learning has emerged as a powerful tool that can analyze vast amounts of data, uncover patterns, and make predictions. While its applications are widely recognized in fields such as healthcare, finance, and entertainment, there has been a growing fascination with its potential in Behaviorism, Cognitive Theory, and Psychoanalytic Theory. We have been exploring the intriguing intersection of deep learning with these mystical domains, delving into the promises, limitations, and ethical considerations associated with the fusion of science and spirituality. Here is some research we are working on in this field.

Machine Learning x Psychology

Illusory Correlation

We use Variational Autoencoders (VAEs) to generate synthetic data that captures the perceived patterns and associations, addressing the issue of illusory correlation

Zeigarnik Effect

To study the Zeigarnik effect, we employ Hidden Markov Models (HMMs) with sequential data, capturing the temporal dependencies in task performance and gaining insights into incomplete tasks and subsequent behavior.

Reactance

We address reactance by using Inverse Reinforcement Learning (IRL) models trained with Reinforcement Learning (RL) algorithms. This allows us to infer individuals' underlying motivations and preferences based on their observed behavior and resistance to external constraints.

Embodied Cognition

We utilize Deep Convolutional Neural Networks (CNNs) combined with Long Short-Term Memory (LSTM) networks to study embodied cognition. This approach enables us to process sensory inputs, such as images, and capture the temporal dynamics of perception, action, and cognition.

Scarcity Effect

We address the scarcity effect by employing Generative Adversarial Networks (GANs) to generate synthetic data. GANs simulate scenarios with limited resource availability, allowing us to investigate the impact of scarcity on decision-making and behavior.

Contrast Effect

To examine the contrast effect, we use Prototypical Networks, a type of Metric Learning algorithm. Prototypical Networks learn a metric space to identify and compare instances that highlight the influence of contextual contrast.

Stereotype Threat

We study stereotype threat by training Multi-Task Learning (MTL) models using deep neural networks. MTL enables the model to simultaneously learn from multiple tasks related to stereotype activation, capturing the interplay between biases in one task and their effects on performance in other tasks.

Peak-End Rule

We address the peak-end rule by utilizing Variational Autoencoders (VAEs) to learn compressed representations of stimuli. This allows us to capture the salient features that contribute to judgments of overall experience, aligning with the concept of peaks and ends influencing subjective evaluations.