As Psychologists, we are often interested in interaction effects that are small in size. Finding these, however, is a challenge, with even the most advanced statistical methods falling foul of multi collinearity, order effects, singularity and model non-convergence. I share my experience of the Random Forest approach - it allows me to enter all my predictors, co-linear and all, deals with non-normal distributions intuitively and can also assist in variable selection if I ask it to.