Christopher Hayes covers a study demonstrating that time in higher security prisons produces increased recidivism. The study linked to is ingenious and persuasive -- it examines continuous offender risk scales that place defendants in prisons of different security levels and finds discontinuously large leaps in recidivism rates at the thresholds that result in higher security placements.
I would like to see someone compare recidivism rates in institutions that don't use as much risk segregation. That might help tease out whether it is the low-risk offenders' association with the high risk offender in high security institutions, or the greater dehumanization and loss of liberty that comes with being placed in a high security environment. If the "low-risk-high-risk" offenders he described (those just above the threshold of the high risk placements) had comparable rates of recidivism in mixed populations but a lower security environment, that might show that it is the association with more hardened offenders, and not the intensity of the prison environment, that is doing the damage.