2022 | Computer Software | Early stage | SDG1 | SDG10 | SDG11 | SDG12 | SDG13 | SDG14 | SDG15 | SDG16 | SDG17 | SDG2 | SDG3 | SDG4 | SDG5 | SDG6 | SDG7 | SDG8 | SDG9 | United States
AI Bias Correction via Machine Learning for Categories

Company or Institution



Computer Software




United States

Sustainable Development Goals (SDGs)

SDG 1: No Poverty

SDG 2: Zero Hunger

SDG 3: Good Health and Well-being

SDG 4: Quality Education

SDG 5: Gender Equality

SDG 6: Clean Water and Sanitation

SDG 7: Affordable and Clean Energy

SDG 8: Decent Work and Economic Growth

SDG 9: Industry, Innovation and Infrastructure

SDG 10: Reduced Inequality

SDG 11: Sustainable Cities and Communities

SDG 12: Responsible Consumption and Production

SDG 13: Climate Action

SDG 14: Life Below Water

SDG 15: Life on Land

SDG 16: Peace and Justice Strong Institutions

SDG 17: Partnerships to achieve the Goal

General description of the AI solution

This proposal specifically addresses a key research problem to better address the needs of data scientists: how to make recommendations on knowledge graphs — which are “category algebras'', aka “categories'' of a special form — in a way that conforms to arbitrary given symbolic logical constraints specified by domain experts. For example, how to take a knowledge graph about cleaning products that may associate Chlorox and Formula 409, but not recommend that pairing for sale because bleach and ammonia are a toxic combination. Currently, data scientists have no such “safety net" and must account for domain rules manually during e.g. feature engineering, a tedious and error-prone process not known to even be possible outside of some particular special cases (e.g. transitive relationships) — a situation we aim to address with this proposal. That is, this research project will draw from the fields of knowledge graphs and graph neural networks in order to evaluate the technical feasibility of (machine) learning “graph embeddings'', that are guaranteed to respect (given) first-order logic constraints — and to apply these new techniques to graph learning projects across domains. In addition to obvious applications in e-tailing for product recommendations, this mechanisms can also be used to explicitly correct bias while learning over graph data.









Public Exposure


International Research Centre
on Artificial Intelligence (IRCAI)
under the auspices of UNESCO 

Jožef Stefan Institute
Jamova cesta 39
SI-1000 Ljubljana



The designations employed and the presentation of material throughout this website do not imply the expression of any opinion whatsoever on the part of UNESCO concerning the legal status of any country, territory, city or area of its authorities, or concerning the delimitation of its frontiers or boundaries.

Design by Ana Fabjan