Herein are just excerpts. Warm up with this, then read the article at your leisure.

Mandel lays out the hindsight problem, unmitigated by the current regime, with those on the KSR side, wishing to gut the suggestion test, full of, at best, wishful thinking.

The hindsight effect is familiar to all—consider the widespread adages “hindsight is 20/20” or “Monday morning quarterback.” These sayings are based on a now well-proven fact: once outcome information is known, people are cognitively incapable of preventing that information from influencing their understanding of past events. As a result, individuals consistently (and unconsciously) exaggerate what could have been anticipated in foresight and not only tend to view what occurred as having been inevitable, but also as having appeared relatively inevitable beforehand.

The human incapacity to ignore hindsight information has been well-studied and documented in a variety of contexts. The hindsight bias has been confirmed in over one-hundred studies of both lay and expert judgment in both laboratory and real world settings in many fields. The hindsight bias has been verified in legal decision as well. Several mock jury studies in the tort and search and seizure contexts demonstrated significant hindsight effects in judging negligence and probable cause, respectively.

The dilemma that the hindsight bias poses for patent law is severe... This hindsight effect creates substantial difficulty for patent law. Judges, jurors, and patent examiners will routinely view inventions that were actually non-obvious at the time of invention as instead having been obvious, because the invention is known to the decision-maker at the time the non-obvious determination is made in hindsight. The first study in the current line of research confirmed this effect. That study of mock jurors charged with evaluating whether a patented invention was obvious concluded, “the hindsight bias significantly influences non-obvious judgments . . . . Ex post knowledge of invention deeply affect[s] participants’ conclusions regarding whether an invention [is] non-obvious ex ante.” The study also found that the magnitude of hindsight bias for non-obvious decisions was greater than that reported for other legal judgments.

The petitioner in KSR, backed by amici briefs from the Solicitor General of the United States, industry representatives, and a number of intellectual property professors, argues that the suggestion test violates the Patent Act and the Supreme Court’s non-obvious precedent because it results in certain actually obvious advances instead being held non-obvious. Their argument is that combining references may have been obvious even if there was no explicit suggestion, teaching, or motivation to combine the references in the prior art.

The experimental study reported here indicates that these contentions are not supported by the reality of human decision-making... The suggestion test does not appear to have the problematic effect asserted by its detractors.

[T]he debate before the Supreme Court over the suggestion test is foundationally misplaced. The debate as currently constituted focuses on whether the suggestion test is overbroad—whether, in an effort to reduce the hindsight bias, the test also precludes decision-makers from considering all appropriate evidence of obviousness. Both supporters and detractors of the suggestion test appear to take as a given that the suggestion requirement does ameliorate the hindsight bias; supporters argue that it does so appropriately, detractors effectively argue that it goes too far. The findings reported here reveal that these base assumptions are false.

The instant study also examined the impact of Supreme Court jurisprudence on the hindsight bias in non-obvious decisions. Under the Supreme Court’s decision in Graham, a judge or jury must follow a carefully structured framework in order to determine whether an invention is non-obvious. 

The hindsight bias thus remains unresolved in current patent jurisprudence.

Mandel then describes experimental studies of the suggestion test and the Graham framework for obviousness evaluation. As to its conclusions -

As expected, participants rated inventions non-obvious significantly more frequently in foresight than in hindsight... The frequency of participants that judged the invention obvious was higher, in both foresight and hindsight conditions, respectively, for those who received a scenario with a suggestion to combine prior art references than for those who had no suggestion in their scenario... The suggestion instruction did not significantly reduce the hindsight bias among mock jurors.

Mock jurors who were instructed to follow the Graham framework in a manner based on model patent jury instructions demonstrated no less hindsight effect than mock jurors who were not so instructed... Graham instructions had no significant effect on judgments of obviousness... [T]he Graham framework does not ameliorate the hindsight bias.

Once an outcome is known, individuals are cognitively incapable of properly discounting it. “[T]he hindsight bias [is] essentially impossible to avoid . . . . correcting for the bias is not feasible.”

PTO examiners may fare better with the non-obvious hindsight bias. A meta-analysis of over one hundred hindsight bias studies found that individuals familiar with a task, either because they have experienced it or because they have expertise in an area relevant to the task, demonstrate slightly less of a hindsight bias than individuals who are unfamiliar with the task... [But,] the benefit of familiarity for the hindsight bias, however, is slight.

As to possible solution -

Given the substantial and now confirmed prejudicial effect of the hindsight bias, ... the suggestion test should be retained for several reasons: it does not appear to cause the harms hypothesized by its critics, it potentially reduces the hindsight bias for complex technology inventions, and the problem the test confronts (erroneous decisions that an invention is obvious in hindsight) is known and significant while the problem the test is alleged to create (over-compensation for the hindsight bias) is unconfirmed and conjectural.

The sole means identified to date to debias non-obvious decisions is to bifurcate the non-obvious issue.

Under the jury trial proposal, pre-trial hearings would be used to determine the problem the inventor was working on or the problem the invention solved, the scope and content of the analogous prior art, and the admissibility of any expert testimony concerning the level of ordinary skill in the art and what a PHOSITA would know how to do. The jury would not be informed of the invention. Trial would be conducted based on the problem the inventor was working on, the analogous prior art, and the skill and knowledge of a person having ordinary skill in the art. The jury would deliberate concerning whether the advances sought were obvious to a person having an ordinary level of skill in the art. By preventing jurors from having hindsight knowledge of the invention in the first instance, the jurors would be placed in the appropriate foresight position to judge whether the invention is non-obvious.

Bifurcation also can be applied at the PTO in the initial patent examination to determine whether to grant a patent. Bifurcation at the PTO would require two examiners.

Bifurcation is not a perfect or complete solution, but will produce far more accurate non-obvious decisions than current methods and jurisprudence. In addition, bifurcation would eliminate the need for the suggestion, teaching, or motivation requirement in the first instance. If there is no hindsight bias then there is no need to mitigate its effect... Though bifurcation is feasible at the PTO and in many jury cases, it is not feasible in certain circumstances, particularly for patent cases tried to judges. In situations where the non-obvious issue cannot be bifurcated, the suggestion requirement should be retained.

Again, this synopsis does not do the article justice; read the article.