By Daria Onitiu, Northumbria University
Setting the scene: discerning ‘fashion’ with AI and data personalisation
I always was an enthusiastic observer of people’s styles. During my daily walk in Newcastle, the United Kingdom, I do recognise my own preferences in clothing, such as asymmetric clothing in dark colours and certain new fashion trends. Once I decide to look up a new style or an outfit online, I am overwhelmed by the choices and possibilities to buy certain items of clothing.
Fashion recommender systems precisely intend to make my styling choices more efficient, seeking to understand costumer behaviour and recommend products that are likely to be purchased. Fashion recommender systems, entailing the use of artificial intelligence on e-commerce websites as well as subscription-based services, support users to filter between a large amount of products, providing users with a personalised shopping experience.
Building robust fashion recommender systems is difficult, due to the seasonality of trends and fragility of consumer preferences. However, over the years and with advancements in neural networks, fashion recommender systems are able to deal with a larger set of attributes, such as garment texture, shape, size, fashionability, and the individual’s unique characteristics and preferences. I call these set of attributes ‘fashion narratives’, because they tend to reflect the interaction between product attributes on the combination of clothing and user attributes, such as user’s perception of style, social context or mood.
Recognising the benefits of fashion recommender systems for consumers, a key question is how fashion narratives in recommender systems interact with an individual’s right to privacy. My research intends to provide for an interdisciplinary outlook of privacy, focusing on the individual’s interaction with personalisation systems in fashion.
How ‘fashion narratives’ are captured by ‘data points’ pertaining to the individual
In order to address the research question above, it is important to understand how fashion recommender systems interact with the individual and to develop an understanding of the meaning of ‘fashion’. Fashion is much more than the clothing an individual is wearing; but also carries cultural, social and emotional connotations. However, fashion recommender systems using a neural network methodology, emphasise the importance of clothing attributes that determine the link to the mental concept of an individual, such as a person’s current mood or dressing occasion. These ‘intelligent fashion recommender systems’, using matching criteria between customer behaviour and product attributes in clothing, raise issues regarding an individual’s autonomy and informational self-determination to reveal one’s fashion choices.
I argue, recognising the cultural, social, and emotional aspects of fashion that fashion recommender systems can cause issues in respect of the right to privacy. The individuals’ identification process is shifted to the interpretation and correlation of data points, outside the control of the user. Hence, what would be needed is an understanding of privacy that addresses the problems of (a) the performative function of fashion identity and (b) perception of knowledge inference, both impacted by fashion recommender systems.
Therefore, the next question is how to articulate the privacy problems outlined above in a way that addresses problems fashion recommender systems have on an individual’s identification process with appearance management and perception. In order to answer the question, further research intends to provide lawyers dealing with privacy problems a toolkit tailored to the notion of ‘fashion identity’.
Understanding ‘privacy’ in light of ‘fashion identity’
An important contribution to a dynamic understanding of privacy in terms of personal autonomy is stipulated in case law by the European Court of Human Rights (ECtHR) on article 8. According to ECtHR case law, the right to privacy is interpreted broadly and the scope of article 8 of the European Convention on Human Rights (ECHR Convention) includes an individual’s development of personality and enter relationships. However, the interpretation of article 8 of the ECHR Convention does not cover situations that have an impact on the subjective sense of self, such as instances where my choices to be a specific ‘type’ of person, would be impacted by the frequent interactions with fashion recommender systems. What does this mean in light of the right to privacy, when a fashion brand gives you an indication on who you are, such as one of a ‘girly’ type? I argue that it is that impact of artificial intelligence on the perception of and inference of knowledge of self that is not covered by our current understanding of privacy.
Identifying the meaning of ‘fashion identity’ can offer the means to shape our understanding of privacy in the digital age to address the risks fashion recommender systems have on the objective and subjective sense of self. ‘Fashion identity’ illustrates the various forms of expression and communication of appearance that situates a subject into a social context. Whilst fashion theory does not offer a clear and definite answer to ‘what is fashion’, analysing the terms ‘dress’ and ‘clothing’ gives a valuable starting point to view ‘fashion’ in its many forms.
Identifying the personal and social aspect of ‘fashion’ intends to clarify the parameters the right to privacy should seek to protect when challenged by these new technologies. It can clarify the parameters of social interaction, focusing on the conditions of identity construction. Connecting the definition of fashion identity with the right to privacy identity’ could clarify the impact of fashion recommender systems on individual behaviour that is shaped by new mediums that negotiate new values.
Fashion recommender systems offer great opportunities for consumers and fashion brands’ to personalise the shopping journey. Significant advancements in neural networks make recommender systems more ‘intelligent’ working with a better detection of product attributes.
Nevertheless, an important part of ‘fashion’ is to enhance an individual’s personal identity. During my daily walks in Newcastle, I am not only an observer, but I make conscious and subconscious choices on the kind of style concerning the management of appearance and perception. That process entailing the exploration of my identity, risks to be gradually shifted to the analysis of ‘data points’ in recommender systems. Lawyers need to pay attention to this development and investigate a nuanced approach to privacy that includes an individual’s perception on the formation of one’s identity. Discussing the personal and social aspect of fashion clarifies how privacy can safeguard the unique nuances of representation and perception of an individual that are exemplified in actual behaviour.
 See for example, LC Wang, XY Zeng, L Koehl, and Y Chen, ‘Intelligent Fashion Recommender System: Fuzzy Logic in Personalized Garment Design’ (2015) 45 (1) IEEE Transactions on Human-Machine Systems 95.
 Leane Luce, Artificial Intelligence for Fashion: How AI is revolutionizing the fashion industry (Apress 2019) 12.
 See for example, Susan J Vincent, ‘Fashioning the Body Today’ in Linda Welters and Abby Lillethun (eds), The Fashion Reader (Oxford: Berg 2011) 358.
 Congying Guan, Shengfeng Qin, Wessie Ling, Guofu Ding, ‘Apparel recommendation system evolution: an empirical review’ (2016) 28 (6) International Journal of Clothing Science And Technology 854, 863.
 Agnes Rocamora and Anneke Smelik, Thinking through Fashion (London: I.B. Tauris 2016) 3; see also, Joanne Entwistle, The Fashioned Body (Blackwell Publishers Ltd 2000) 40-41.
 Joanne Entwistle, ‘Introduction’ in Sandy Black, Amy de la Haye, Joanne Entwistle, Agnes Rocamora, Regina A Root and Helen Thomas (eds), The Handbook of Fashion Studies (Bloomsbury 2013) 97.
This post was originally due to be presented as part of the IT Law and Cyberspace stream at the 2020 Annual SLSA Conference, due to take place at Portsmouth University.