AI Algorithm as Personal Style Tracker

On: September 24, 2017
Print Friendly, PDF & Email
About Vera Seegers


Design specifically with a GAN, to overcome fast-fashion and overconsumption

Anya Hindmarch

What will happen when designers can exactly produce what consumers want and need? The generative adversarial network (GAN) founded by online shopping company Amazon can actually do this by tracking, reacting and shaping the latest fashion trends. Designers can from now on design more specifically, so companies can completely dominate the retail field (Knight,1). But even better, this tool can also maybe a solution to overcome issues as fast-fashion and overconsumption.

The GAN as style tracker

Every single person is unique and differs from others. A human being is having a personal character with specific style preferences to design his own life. But the way he is shaping his own identity is highly influenced today by the fast-changing given trends of our society. The dynamic of the increasing influence of social lifestyle tendencies is penetrating our lives both online (social media) as offline (urban marketing) ceaselessly (Sassen,49). These personal preferences are a mix of uniqueness, conformation, and culture, who are shaping the abstract consumer (Kim & Markus,785) These preferences can be transformed into useful data for designers to create styles that are more tailored to an individual. This is currently being made possible by an AI Algorithm designed by a cutting-edge tool called a generative adversarial network (GAN). Online shopping company Amazon came up with this idea to get initially more grip on the fashion industry and to try to dominate the field. The GAN system is using raw data and is internalizing the properties of a particular style, simply by looking at lots of examples. After this analyzing process, the GAN system can then apply that specific founded style to an existing or future product, for example, fashion and furniture to provide more specific retail-making recommendations (Knight,1).

“The GAN is basically training the algorithm by analyzing the data that comes out of a few labels (for example hashtags) who are attached to the posted images on social media platforms as Instagram and Pinterest. Usually the algorithm is just requiring extensive labels, but in this case, the algorithm is getting smarter from this visual information and is using it to inform designers very precisely to innovate new products. It’s basically learning from a particular style by analysing images, and then the system generates new items in a similar style.” (Knight,1)

Can the GAN overcome over-consumption and fast-fashion?

The fashion industry is nowadays one of the most terrible waisting industries, which is a huge problem for our environment (Cachon, 778). The fast fashion system reacts fast on short “hot” fashion trends by producing cheap and fast monthly and even weekly collections, which is turning into over-consumption. According to Gêrard P. Cachon, this system has arisen because of the uncertain preferences of consumers. Fashion companies want to make much money as possible, and the production of so much (trash) fashion and throwing it away after the season, is making that huge profit nowadays. Though it is really bad for the environment until now there was no other option to satisfy the consumer (Cachon, 778-784). By using this GAN system for designing products, it will maybe possible to overcome fast-fashion and thus over-consumption, because designers from brands and companies can design now exactly what people want and need.

Several attempts have already been made to overcome fast-fashion. An example is the James Dyson award-winning origami children’s clothing (Petit Pli) from fashion designer Ryan Yasin. He combined airplane technology with the 3D printer capabilities. The result is long-lasting clothing that grows up to 3 years with the child. In this case, the fabric and the design are sustainable, but the designer still has to know in which particular style he has to design. (Click here to see his collection)

(Click here to see his collection)

The problem nowadays is also that (fashion) trends are changing so fast because of the online culture which got dominated by bloggers and bloggers on social media platforms as Instagram and Pinterest. They produce and decide the tendencies for the leading trends by posting look photographs online. The current research methods of fashion forecasting cannot keep pace with the dynamics of this constantly fast-changing (online) marketplace wherein this huge amount of available data is leading (Rickman & Cosenza, 609). The challenge is to tap the continual flow of data/information from the present, to contrast it with the already stored set of information coming from the past, to create the best-resulted insights into specific trends (Rickman & Cosenza, 614). The ‘Nielsen BuzzMetrics’ and the ‘BlogPulse’ system have already tried to get more grip on fast-changing fashion trends, but it did not work good enough to get precise results (Rickman & Cosenza, 602). It seems like that by tracking online influences and behavior with the GAN system, designers and companies can easily anticipate on it, because it is giving the fastest and precise insights in this even faster changing online fashion culture.

The GAN and the “filter bubble”

The GAN system is mainly using raw data which comes from social lifestyle networks as mentioned before: Instagram and Pinterest. It is working by two main deep integrated networks which are operating in a tandem. It is a new innovative form of communicating and networking with consumers. This GAN system is so specific in tracking, shaping and reacting that it is creating a so-called “Filter Bubble.” Internet activist Eli Pariser came up with this term in 2011. Pariser, Bill Gates and many others think that the “Filter Bubble” will be a huge problem. Filter Bubbles arise from personalized searches wherein a specific website algorithm selects out of these searches and interests what that specific user wants to see. The result is that people will get isolated in their own cultural and ideological bubble (Pariser,7-15). The GAN is also doing this because by using the data which is produced by a specific user, so the filter bubble is producing a personal ecosystem of information by using the algorithm. The products that will arise are coming out of this ideological close framework. There will be no space for other influences or ideas (in this case about fashion). On the one hand, this is positive for the companies to get grip on the fashion field and their consumers. Also, the fast fashion and overconsumption problems will be solved because of these specific outcomes of data, which will be amazing! But on the other hand consumers will always stay in their own “comfort zone” which is not specifically a bad thing, but fashion can turn into a boring thing. And in the end, people will buy less because over the time they will get faster satisfied in their own “bubble.”


The GAN system is a positive and good way to dominate the fashion industry. Companies can make the profit in a sustainable way by design specifically, which is for both parties (the companies and the environment) a long-lasting sustainable economy. Fast fashion does not have to exist anymore in this way, because the consumer will get satisfied by specifically designing, so overconsumption will be avoided. Also, the idea of children’s fashion designer Ryan Yasin is a sustainable way because of the stretchable fabric he designed. But still, people want to have a specific design in their own specific style. So when Ryan can use this GAN system in combination with his stretchable fabric, I think that it will be the ultimate way to produce a long-lasting, tailored design clothing line. On the other hand is this GAN system producing a “filter bubble” for every specific user. Eli Pariser thinks that this will be a bad thing. I think that in this case, the filter bubble is more a good thing than a bad thing. Producing specifically is now the only option to avoid fast fashion and thus overconsumption, and on the same time to let companies make a profit. The “filter bubble” got created by this system, but it definitely has to be there to make this GAN system work. One thing that can happen, is that consumers will get faster satisfied in their own style or trend because there is no new input that will enter their “bubble.” Styles and trends got produced by people. This generally occurs on the aforementioned social media platforms as Instagram and Pinterest by the so-called bloggers and vloggers. They are the ones today who make the decisions in fashion indirectly, by posting photographs about their looks. They are the inventors of a specific style or trend. The next big step in this process will probably be that the computer is generating a completely new style, which is really resonated with people. This can then easily be forced to consumers, because of the arising of the “filter bubble.”


Cachon, Gêrard P.  “The Value of Fast Fashion: Quick Response, Enhanced Design, and
Strategic Consumer Behaviour.” Management Science Vol 57. Part 4. (2011): 795.

Dezeen, “Ryan Mario Yasin’s Petit Pli kids clothing expands to fit as children grow.” Youtube.
07 Juli 2017. September 18, 2017.

Hindmarch, Anya. Colourful Iluminated Surfaces. 2016. September 19, 2017.

Kim. H, & Markus, H.R. “Deviance or uniqueness, harmony or conformity? A cultural
analysis. Journal of Personality and Social Psychology Vol 77. Part 4. 785-800.

Knight, Will. “Amazon Has Developed an AI Fashion Desginer.” Technology Review
Part 1 (2017): 4. August 12, 2017.

Pariser, Eli. The Filter Bubble: What The Internet is Hiding From You. London, England:
Pinguin Books, 2011.

Rickman, Tracy Anna, Cosenza, Robert M. “The changing digital dynamics of multichannel  marketing:
The fesibility of the weblog:  text mining approach for fast fashion trending.” Journal of Fashion
Marketing and Management 
Vol 11. Part 4. 604-621.

Sassen, Saskia. “Relational Space.” Urban Screens Reader. McQuire. S: Martin. M: Niederer,
S.M.C. Amsterdam: Institute of Network Cultures, 2009. 288.


Comments are closed.