In this installment, we’ll delve into a key analytic that uncovers the insights behind these 100 images: Subcultures and Aesthetics. We’ll explore how they help us decode consumer experiences and discuss the specialized classification model we’ve engineered at Quilt.AI.

Subcultures and Aesthetics typically draw from diverse sources of human creativity — fashion, music, art movements, and architecture. Historically, they’ve been more than just superficial markers, often signaling the values and beliefs of respective communities.

Over the past five years, we’ve observed a shift in how subcultures and aesthetics influence identity formation — gone are the days of enduring “goth phases” akin to today’s fleeting “tomato girl summer.” Despite their reduced permanence in mainstream culture, their impact persists in a different form. For digital natives, these categories act as platforms for exploring and consuming new philosophies, ideals, and styles.

Nevertheless, the relevance of subcultures and aesthetics persists, particularly in spotlighting emerging trends and attitudes. For example, current cooking approaches might align with ‘Lazy Girl Dinner’, focusing on quick and easy recipes for the busy, overworked professional, or the ‘Cottagecore’ ethos, which romanticizes the ideals of making everything from scratch. This helps us analyze experiences quickly and in depth on Culture Lens.

The dictionary features mood boards and comprehensive descriptions that encapsulate the essence of each category.

In the process of building and updating the model, challenges such as class imbalance do arise. Less popular subcultures often get miscategorized. We tackle this issue through collaborating with cultural researchers to ascertain whether frequently miscategorized subcultures and aesthetics can be simplified or better differentiated in the training data. During this data cleaning process, we aim to emphasize distinctions between similar subcultures and aesthetics in our dictionary. If clear differences are lacking, we remove the less popular options.

We then fine-tune the model using a combination of feature selection and dimensionality reduction methods. This meticulous approach ensures the highest level of accuracy in our model. This empowers researchers to explore consumer experiences with greater depth and nuance through the lens of subcultures and aesthetics.