Description as a Tweet:

You've just gotten rejected for a loan by an ALGORITHM! What should you do? Ask DiCE!

Inspiration:

I do research in counterfactual explanations. I also needed to learn to use DiCE and Dash for unrelated projects. This was a learn-a-thon.

What it does:

Machine learning models make decisions with real impact on individuals every day. Given an unfavorable automated outcome, what can an individual do to change it? For example, if you are automatically rejected for a loan, how can you be accepted? This field of research is called Counterfactual Explanations.
Microsoft released DiCE (Diverse Counterfactual Explanations) to answer this question. This webpage is a simple demo of DiCE on the Adult Income Dataset. It allows users to play with the dataset and counterfactual explanations given by DiCE.
It was created using plotly dash. The target model is a random forest built with sklearn.

How we built it:

I used:
1. Dash (data-driven python web framework)
2. Dash Bootstrap
3. DiCE
4. SKlearn

Technologies we used:

• Python
• AI/Machine Learning

Challenges we ran into:

I worked alone because the project is so specific and closely connected to my research.
I got stuck on many front-end elements.

Accomplishments we're proud of:

I learned how to use Dash and DiCE in a short time.

What we've learned:

Dash and DiCE

What's next:

I use what I learned in other projects.

Built with:

See the previous question about implementation details.

Prizes we're going for:

• Best AI/ML Hack

Jake Valladares

Table 7