At Sentencing Law & Policy, Professor Berman points to Judge Posner's discussion of 3553(a) factors in a child pornography case. Judge Posner points out that 3553(a) factors seemingly require empirical evidence to be introduced by the defense:
The defendant failed to give the district judge any help in picking a sentence. It was not nearly enough for his lawyer to point out that his client had been a victim of child molestation; the lawyer presented no evidence or studies to indicate that such a history makes a person less able to avoid becoming a child molester, let alone becoming a producer of child pornography (which might however be considered a form of child molestation when it involves photographing actual children in erotic poses).
I think this is particularly true in lots of sex offense cases where judges take for granted that all sex offenders have a very high risk of recidivism even though empirical evidence does not support the evidence often cited in court opinions. However, I think it is a tough burden for most public defenders and court-appointed attorneys to gather the empirical evidence to persuade a sentencing judge. Judge Posner's 3553(a) reasoning sets a high bar that I would guess the overwhelmingly majority of defendant's wouldn't be able to meet even if they had a very good 3553(a) argument.