Question regarding interpretation in Chapter 22

[jfklahdgha]

Hi @matheusfacure, thanks for this awesome book!

I am currently reading Chapter 22, and am curious about the interpretation for the non-parametric double machine learning here: https://matheusfacure.github.io/python-causality-handbook/22-Debiased-Orthogonal-Machine-Learning.html#what-is-non-parametric-about.

Here it says the non-parametric method finds "the slopes of the lines that are tangential to the outcome function at the treatment point". However, in the fitting procedure, we are fitting \tau(X) to be Y_residual / T_residual, so it is more like a ratio of two residuals, rather than their slope (d Y_residual / d T_residual). They are equivalent only when there is no intercept term. I wonder did I miss anything here?

Thanks!