Human recognition and human-computer interaction

Some piece of news sometime in the past years said that recognition of human emotions etc by computer is practised with actors. That is just insane: acting looks different from really doing. Of course human expressions, body language, looks and tones of voice & typical reactions should be copied from real people doing real thinsg and not just from actors with who-knows-what real intentions. What one could start from is basic things humans do like watching, listening, moving and practising sports, doing practical things. And in those there are easier parts and more difficult parts, successes and failures, so that one could learn such features from those - or from ordinary photograpfs. Then one could also understand differencies in ways of thinking: ways of using the sense of sight, remembering, going things through. If thing is done exactkly the same way, it gives the same result on anyone doing, I guess, so one does not need to understand each person separately but basic features well enough, very basic enough and that's it: all human endeavours, all ways of doing gone through objectively and detailedly.