Skip to main content
#1AI in practice

Feeding Information for Sustainability The Green Consumption Assistant

Magazin #1 | Sommer 2022

Feeding Information for Sustainability: The Green Consumption Assistant

Not having any sustainable options when buying a product online should no longer be an excuse for unsustainable consumption choices. The Green Consumption Assistant project sets out to support consumers in easily finding and buying sustain- able products – by making use of
the existing machine learning infra- structures in the retail industry.

The fundamental dilemma observed by the project team around Tilman Santarius and Maike Gossen from TU Berlin, Felix Biessmann from the Berliner Hoch­ schule für Technik and the green search engine Ecosia was twofold. First, people say they want to make more sustainable choices but do not act on that desire when buying products. Second, the existing machine learning tools in the retail industry could be used to make sus­tainable consumption decisions a lot easier, but there is a lack of essential and comprehensive data about sus­tainable products to feed these systems.

The solution: Building green databases and making sus­ tainability aspects an essential criterium for an algorithm’s automated decision­making. In the sphere of online shopping, automated recommender systems could then rank sus­ tainable products more promi­ nently than non­sustainable products. Transparency about such databases would likewise allow for systematic checks of what sustainability definitions and certifications are being used as the basis for a prod­ uct’s labeling as sustainable. That would ultimately empower consumers to make more in­ formed choices.

The Green Consumption As­sistant addresses exactly this lack of green databases. The project team has been working on creating the GreenDB, a database containing sustain­ ability information for consumer goods. The GreenDB is up­ dated on a weekly basis and includes over 220,000 unique products from the largest online retailers in several European countries. In contrast to previous approaches to sustainabil­ ity databases, the GreenDB covers only products that users are interested in: Its 26 product categories, currently mostly fashion and electronics, have been selected based on a care­ ful analysis of the search logs of Ecosia users. The database displays information on the type of sustainability information that underlies any given product in the database – be it either more credible third-party verification or non-verified private sustainability labels. The database is used in the shopping tab of Ecosia’s search site highlighting green products and thereby possibly encouraging consumers to make more sustainable choices. The GreenDB is publicly available and has two main purposes. One is for research. The other is for improving AI applications, such as recommendations and the reliability of sustainabilty information.

Beyond the usage of the GreenDB to drive sustainable consumption, it can also help to gain new insights into the availability of sustainable infor­ mation on online fashion retail – and potentially infer appropriate policy changes. The relatively small ratio of only 14 percent of sustainability­ tagged products in the online shops of Germany’s largest fashion retail­ ers are labeled with credible third­ party verified sustainability labels. This underlines the difficulty faced by consumers in determining how sustainable a product is. The widely used private and non-certified labels prevent comparability and add con­ fusion and uncertainty for consum­ ers. More clarity and information are urgently needed, especially political initiatives tackling the risk of green­ washing resulting from uncertified and weak sustainability information.


sustainability potential in application


economic sustainability


sustainability potential in application


promotion of sustainable products

AI systems can be deployed in online shopping to promote more sustainable consumption through recommendation and search algorithms. Sustainable products can be given greater visibility in listings, product search results can highlight more sustainable alternatives and additional information can also be displayed, such as CO2 emissions. AI systems should promote economization and sufficiency and encourage a shift away from unsustainable patterns of use (such as binge watching and food waste). Sustainability criteria like CO2 emissions, working conditions and fairness must be programmed as relevant criteria in the decision-making process of the systems.

Green Consumption Assistant

The Green Consumption Assistant (GCA) helps consumers in making online purchasing decisions that are more sustainable. It displays green product alternatives on the Ecosia search engine and provides information about more sustainable alternatives, such as references to repair, rental or sharing options. The basis for GCA’s recommendations is a product database (GreenDB) of environmental and social sustainability information developed with the help of Machine Learning.

The GCA project is a partnership between the Technical University Berlin, the Berliner Hochschule für Technik and the green search engine Ecosia. This model project on the use of Artificial Intelligence in addressing environmental challenges has been funded by the German Environment Ministry.