A mainstream assumption in psychiatry is that DSM revisions that it can be, or are being, gradually improved with every iteration. In contrast, critics think DSM is radically flawed, and seem to support a wholesale erasure of it, and a new nosology starting from zero. The mainstream view has its liberals and its conservatives; liberals want to tinker more, conservatives less. Both are pragmatic, but DSM-liberals will bow to sufficient scientific evidence over purely pragmatic concerns, while DSM-conservatives take pragmatism to its very end - the postmodernist aversion to science when it conflicts with their wishes. So mainstream DSM liberals want to make small to medium changes, mainstream postmodern conservatives no changes. Non-mainstream critics include those who disbelieve in any nosology, and thus would not have any suggestions for a new and improved version. A final grouping among the non-mainstream critics of DSM are those who accept the reality of a few - very few - mental illnesses, but who feel that they are rare, and need little medical intervention. (This group may include libertarian critics like Thomas Szasz and social constructionists like David Healy).
In my view, the question of evolution versus revolution doesn't even arise if we keep our nosology on the current foundation given by DSM-IV - postmodernist pragmatism. This approach is stagnant; it neither progresses nor retreats; in the absence of any notion of progress, the question of whether change is to be slow or fast is irrelevant.
So the first step, I believe, is to explicitly reject postmodernist pragmatism, which means going against the powerful conservative leaders of DSM-IV (the last revision in 1994), who vociferously oppose many of the changes in DSM-5.
Then the evolution versus revolution question can be asked.
If one takes the realist perspective at all - if the reality of any mental illness is allowed (even if only in theory) -then the avowedly postmodernist rejection of all nosology is not an option. This leaves mainstream DSM liberalism and its social constructionist critics. I think they both have some effective points, and major weaknesses, so that my proposed solution would involve some points of agreement with each group.
DSM liberals, many of whom head the DSM-5 process, appropriately value science. The criteria for change are scientific: studies of nosology with the classic criteria of symptoms, course, genetics, and treatment response, augmented by biological markers, pathophysiology, and other biological data as they become available. Where such scientific evidence does not support a biological disease-entity, then the social constructionist view is more likely to be valid. At the same time, anthropological and cultural research can be used to identify and support social constructionist explanations where the biological approach demonstrates the absence of a disease process.
My view here is method-based (in the tradition of Karl Jaspers) - not eclectic nor dogmatic. The postmodern eclectic mix of arbitrary preferences has proven its barrenness. But dogmatic responses aren't helpful. Unfortunately some psychiatrists dogmatically try to fit all clinical conditions into the biological disease model, even when research doesn't objectively support that view. Just as unfortunately, most social constructionists dogmatically try to deconstruct biologically validated diseases like manic-depressive illness, based on the social and cultural contexts of those diseases.
We need to be clear when we are faced with diseases, and when not. Mixing and matching pragmatically will not do; taking only one approach is hardly better. Real scientific work, conducted honestly in awareness of this distinction, will allow us to classify what is what.
Based on these considerations, I think some of the more revolutionary changes in DSM-5 that are based on empirical science - like the introduction of dimensions for personality and the removal of archaic scientifically unproven constructs like "narcissistic" or "histrionic" or "dependent" personality - are advances. Smaller evolutionary changes make sense elsewhere, as in some of the revisions for defining bipolar disorder (e.g., dropping the exclusion for antidepressant-induced mania, based on extensive empirical data). Unfortunately, some major revisions are made without much scientific basis, such as the pediatric "temper dysregulation disorder", introduced mostly on pragmatic grounds to discourage the diagnosis of childhood bipolar disorder. Here conservatism about revision would seem more valid. And other revolutionary changes, such as narrowing the wildly broad definition of "major depressive disorder", aren't even considered, despite important relevant scientific evidence.
Following the method-based approach to psychiatry, and prioritizing of science over pragmatism, I think our approach to DSM revision should be both revolutionary and evolutionary, depending on the condition.