This work was done during one weekend by research workshop participants and does not represent the work of Apart Research.
ApartSprints
ScaleOversight
Accepted at the 
ScaleOversight
 research sprint on 
February 16, 2023

Sustainable Fashion Brand Language Learning Model 1

Problem: Current language learning models cannot determine whether fashion brands are sustainable. Solution: Train a language learning model to accurately determine whether fashion brands are sustainable. AI Safety Topics: Moral Decision Making, Sandwiching Paradigm, Human-AI Moral Alignment Research Objective: Create a morally aligned language learning model that can determine whether fashion brands are sustainable based off of company data and use sandwiching methods to improve accuracy of the model. Current Prototype: A chatbot that can determine whether a brand is sustainable based on given context (sustainability definition, company background) Tools Used: Python, Interactive Composition Explorer, OpenAI API Future Chatbot Goal: A chatbot that can determine whether any fashion brand is sustainable. The model has access to data regarding each fashion brand’s supply chain logistics, wages, and sustainability actions. The user inputs the company name and the model opens up the file with the company data and determines whether the brand is sustainable. To improve the accuracy of the model, it will use the sandwiching paradigm: the model will make initial assumptions about sustainability based on given data and context. Then a group of non-experts will determine company sustainability based on their morals with the same given context to improve the model. Then a group of sustainability experts will determine company sustainability based on their morals with their expert knowledge to improve the model.

By 
Zaina Shaik
🏆 
4th place
3rd place
2nd place
1st place
 by peer review