It is hard to come up with cases where anti-aliasing matters in quantitative image analysis pipelines, because small details (that anti-aliasing artifacts can interfere with) are mostly lost anyway due to decreased resolution of the downsampled image. This image from Thevenaz demonstrates it very well:
Anti-aliasing may matter in very special cases, for example when you are not interested in structural details but just want to preserve overall texture appearance. Probably we don’t need to make the regular imaging pipeline more complicated because of such narrow use cases, but we could add a simple scripted module that offers anti-aliased resampling. If there are any SlicerMorph modules that offer image downsampling, then it should use vtkImageResize (which is, based on its documentation, is very fast and has the best memory efficiency among all the algorithms mentioned above).