When a positive outcome is almost certain, there's less need for fine measurements. For example, how much effort is it worth to determine whether tech job placement for Howard University students goes up by 2x or exactly 2.15x?
For example, Google could just simply give internships to those at HBCU's, an action one may see as definitely positive. However, if for whatever reason those students fail to perform, in the eyes of the non black colleagues, stereotypes of incompetence will be created, if not exacerbated. Black people are punished more harshly for signs of incompetence of similar magnitude than Whites [2].
A somewhat similar thing happened with minority lawyers, as law schools tried to prop up the supply with affirmative action policies [1].
Things are just complicated, in general.
[1] http://blogs.wsj.com/law/2007/08/29/is-affirmative-action-at...
[2] http://www.huffingtonpost.com/2015/05/27/black-students-puni...
This is a straw man. Google isn't doing that. Google and Howard are providing instruction.
Gonna have to strongly disagree, here. Declaring that a "positive" outcome is "almost certain" is an invitation for punting on specifically defining what that "positive" outcome is and then hand-waving any measurement of progress. This is especially dangerous when the "positive" outcome comes at non-trivial expense.
Edit: a word