Counterfactual for subpopulation

[tonyabracadabra]

Can DiCE be extended for calculating the counterfactuals for a subpopulation based on certain conditions?

[amit-sharma]

yes, you can provide a set of rows from a dataframe to the generate_counterfactuals method. One way is to filter the dataframe based on given conditions and then supply the filtered dataframe to generate_counterfactuals.

However, you will still get counterfactuals per row. Is that the goal?

[tonyabracadabra]

yes, you can provide a set of rows from a dataframe to the generate_counterfactuals method. One way is to filter the dataframe based on given conditions and then supply the filtered dataframe to generate_counterfactuals.

However, you will still get counterfactuals per row. Is that the goal?

I was thinking if there is a more efficient way of doing this, I was reading this https://arxiv.org/pdf/2103.11972.pdf and wonder if there is anything similar can help me achieving the contextual algorithm recourse

[amit-sharma]

The method from the arxiv paper is not currently implemented. Will be a great contribution if you/someone is up for it.