Boston Consulting Group (BCG) has released its first open-source software library for human-explainable artificial intelligence (AI). BCG GAMMA FACET enables users to make better business decisions by opening the “black box” of advanced machine learning models. Advances in AI have given data scientists powerful tools to analyze complex business problems and predict outcomes. FACET goes one step further, giving data scientists and business experts a new way to understand how a model arrives at predictions. With this new insight, data scientists can use machine learning models to inform decisions that save money, maximize yield, retain customers, remove bias, and improve patient outcomes.
BCG believes that humans must always be at the core of all AI-based decisions. By helping developers and business users understand how algorithms analyze the data sets on which AI predictions are based, BCG GAMMA FACET reestablishes human control over and trust in AI. It uses a newly developed model-inspection algorithm to explain the relationships between the model variables. And it applies a simulation approach to enable data scientists to conduct “virtual experiments” to determine how changes in these variables can affect predicted outcomes.
“Data scientists are often under pressure to explain the behavior of their models. This is precisely the aim of FACET: to explain the key variables in the models very quickly, in order to provide greater clarity in the dialogue between data scientists and operational teams. By facilitating the explicability of models, FACET contributes to the deployment of a more transparent and more responsible AI,” says Sylvain Duranton, BCG managing director, senior partner, and global leader of BCG GAMMA.
“We are very glad that scikit-learn’s simple and consistent design allowed BCG to develop FACET, a very valuable tool for our community,” says Alexandre Gramfort, senior research scientist at Inria, co-author, and member of the scikit-learn technical committee.
“BCG GAMMA is very excited to join the open-source data science community with our public release of FACET,” says Jan Ittner, BCG partner, associate director, and leader of the BCG GAMMA FACET team. “We look forward to working with the data science community and in partnership with scikit-learn to make AI more useful and understandable for everyone.”
BCG GAMMA FACET is an intuitive, easy-to-implement, open-source software library available to the global data science community.