In K-12 education in Europe, lowest-performing students are given a lot of attention (with "special" classes etc), low-performing get the bulk, and little attention is generally given to high performers. The effort distribution is roughly normal, biased toward lower-end because of people with disabilities.
Goal seems to be to bring people to a reasonable average, but not above it. Education effort seems logical (the bell curve). Inmate "subsidy" does not seem as logical - perhaps a more bell-shaped subsidy distribution would increase the average/mean and also reduce recidivism (if in context of criminal law).
A reasonable imementation of the current argument is to gradually "normalize" the subsidy distribution into bell shape, which will move the "target average" threshold higher along the way.
Generally "normalizing" subsidies along the range of any metric theoretically sounds right. Sadly, given that value of any one person's life (and possible accomplishments) is not defined (=infinite) science cannot provide a solution. F.e. there is more benefit to aociety to help people with minor disabities, then with major ones, but people feel "inhuman" about supporting such statements and decisions this entails. This will become harder still when AI will need to choose allocation of effort to support/save lives.