From Personal Style to AI-Mediated Visual Recommendation

For many years, I approached personal style primarily through practice—fashion media, brand strategy, visual communication, and digital product development. Across these different contexts, I repeatedly encountered the same underlying question: how do visual systems influence the way women understand, present, and negotiate their identities?

Clothing, color, and style are often treated as matters of taste. Yet they also shape confidence, social perception, and the choices women make in different situations. As recommendation systems become increasingly personalized and AI-mediated, these questions extend beyond aesthetics. They become questions of interpretation, trust, explainability, and user autonomy.

My interest in these questions developed through work across fashion media, brand strategy, and digital product development. In each field, visual identity was never simply about selecting what looked attractive. It involved translating bodies, preferences, cultural references, and social expectations into systems that could guide decisions.

Over time, I became increasingly aware of the limits of conventional style categories. Many systems present identity as fixed, reduce complex bodies to simplified types, or offer recommendations without explaining how they were reached. They may help users choose, but they do not always help users understand—or retain control over—the choices being made.

This tension led me to treat personal style not only as a design problem, but also as a research question: how might a visual recommendation system support self-understanding without defining the user too narrowly?

AIFFD emerged from this inquiry as an evolving research prototype for personalized style and visual recommendation. It brings together observations about body structure, facial features, color, personal preferences, and social context to explore how recommendations might become more relevant to the individual.

The prototype is not intended to produce a single, permanent definition of a person’s style. Instead, it provides a structured environment for examining how different forms of information are interpreted, combined, and communicated. It also allows me to investigate when personalization feels supportive—and when it risks becoming reductive or overly prescriptive.

At this stage, AIFFD is as much a method of asking questions as it is a system for generating recommendations. Its development offers a practical way to study how users respond to visual classifications, how explanations affect trust, and how people can retain meaningful agency when working with AI-assisted tools.

Several connected questions guide the ongoing development of AIFFD and the wider inquiry documented through OhSammi:

• Personalization: What information should a visual recommendation system consider, and how can it respond to individual differences without reducing people to rigid categories?

• Explainability: How can a system communicate why it has produced a particular recommendation in language that users can understand and evaluate?

• Trust: What makes an AI-mediated recommendation feel credible, appropriate, and sensitive to the user’s body, identity, and social context?

• User autonomy: How can a system offer meaningful guidance while preserving the user’s ability to question, adjust, reject, or reinterpret its suggestions?

These questions shift the focus from whether an AI system can generate recommendations to how those recommendations are experienced. The quality of a system depends not only on its outputs, but also on whether users understand its reasoning, recognize themselves within it, and remain active participants in the decision-making process.

OhSammi Research Journal is the public record of this evolving inquiry. It provides a space to document not only polished outcomes, but also the questions, assumptions, methods, prototypes, revisions, and uncertainties that shape the research process.

Future entries will examine the conceptual framework behind AIFFD, the development of its body, color, and style systems, and the design decisions involved in translating these ideas into an interactive experience. I will also reflect on broader issues surrounding AI-mediated visual recommendation, including cultural interpretation, representation, transparency, and the relationship between guidance and control.

By documenting the process openly, I hope to make the work more accountable and open to dialogue. OhSammi is therefore not simply a record of what has been built; it is a place to examine how the research changes through testing, reflection, and exchange with others.

This first note marks the beginning of a more systematic research journey. AIFFD will continue to evolve, and the questions guiding it will become more precise as the prototype is developed and tested.

I welcome dialogue with researchers, designers, technologists, and practitioners interested in women’s visual identity, human-centered AI, personalization, and user agency. OhSammi will remain a space for sharing that process as it unfolds.