Could Artificial intelligence help you buy sustainable clothing?
04-12-19 / Joanneke Meijer
Since the horrible Dhaka collapse in 2013, the situation concerning working conditions has somewhat improved and clothing brands are trying to make a change. However a lot more can and should be accomplished. The modern fast-fashion industry is still dominated by the pressure to produce fast and cheap. Every time you buy a piece of clothing, you can make the decision to help the people who produce your clothes, or you basically force them into slavery. If you wear clothes, you are part of the solution!
Many great initiatives are trying to push producers into the right direction. But how do you decide and know which brand is trying to improve the situation? Do you rely on some weirdo on Instagram (yeah that’s me), a sustainable blogger or one (data) source? I always ask clothing stores about sustainability and buy accordingly. Throughout this investigation I unfortunately encountered some untruthful clothing stores. Wouldn’t it be great to have one website you can consult to find one, reliable answer instead? A website that is based on all the information that is out on the web?
Of course, clothing has an impact in many other areas than just human rights, such as environment, material, recycling etc. We could have an ongoing discussion on the exact definition of sustainability or more generally speaking ‘wrong and right’, because it is complicated. One fabric might be fast growing but needs horrible chemicals to be transformed into clothes. Something that might be good for one, is terrible for others. We do not strive for an unbreakable solution (yet), but try to improve the clothing situation step by step.
We want to build one data source containing (almost) all clothing brands with a sustainability rating. This dataset will be publicly accessible to everyone, and could be supplemented with different types of information: prices, countries of production etc. A customer even has the opportunity to add a clothing brand which will be ranked on the spot.
In order to build this data source, we start by making a list of the clothing brands and their corresponding websites. Instead of manually checking sustainability criteria of each brand (working conditions, used materials, recycling policy etc.), we use the available information online. We incorporate the information from both the brand (website) itself and the small set of sustainable benchmark information already available. We use the small dataset to train an Artificial Intelligence (AI) model which will provide sustainability labels for brands that are not yet rated.
For each of these brands, we automatically read (scrape) its homepage and subpages for useful information. This means our model relies on the content a clothing brand provides on their own website. Quality of these pages and the ability to adjust their website to score higher is one of the caveats we are aware of. We encourage every brand to become more transparent and provide more information. Since our model will be self-learning and dynamic, it will adjust the explainable sustainability criteria as brands provide more and better information. Moreover, we believe that brands bluntly providing wrong information on their websites, will (eventually) be caught by the press. We then could be able to incorporate the news into our model and decrease the sustainability ranking of that specific brand.
Using this method for rating thousands of clothing brands will provide a database which has as many brands as you can think of, updated as often as you wish and is adaptive over time. Our open source system will provide well-explainable and reliable answers regarding the level of sustainability of each specific brand.
How we built the model? You can read it in our next blog: How we built the model to help you make a sustainable choice
Questions about the project? Contact Joanneke Meijer.
06 12 62