In his attempt to form a new philosophy for science, Bhaskar distinguishes three possible approaches. One is that of "empiricist realism," spiritually allied to positivists like Comte and Mach. A second is that of "transcendental idealism," a term coined by Kant and which Bhaskar applies to all post-Kantian views of science. Bhaskar, an advocate of a system he terms "transcendental realism," is sympathetic to Kant's attempt to get beyond the limitations of the empiricists but believes that the Kantians tended to subsume all of nature in terms of human perception.
Both transcendental realism and transcendental idealism reject the empiricist account of science, according to which its valid content is exhausted by atomistic facts and their conjunctions. Both agree that there could be no knowledge without the social activity of science. They disagree over whether in this case there would be no nature also. Transcendental realism argues that it is necessary to assume for the intelligibility of science that the order in nature discovered in nature exists independently of man. Transcendental idealism maintains that this order is actually imposed by men in their cognitive activity." -- p. 27.
But is that what Kant himself actually believed? Not so, according to Skidelsky. As a preface to dealing with the philosophy of Cassirer, he provides first a grounding of the German Marburg School in which Cassirer was schooled.
Kant himself had defended Newtonian physics against the skepticism of Hume by grounding it in what he called "pure consciousness..."
The problem, according to Skidelsky, was that subsequence advances in science could no longer be subsumed under Kantian concepts, causing neo-Kantians-- not Kant himself-- to view the order of science to be one "imposed by men:"
[In the Marburg school] The object-constituting function is transferred from Kant's transcendental subject to the evolving practices of physics.And:
Thus could the Marburg school reconcile the nineteenth century recognition of radical historical variety with the eighteenth century ideal of absolute truth.
Whether I will come to agree with Skidelsky that Cassirer's Kantianism is similarly conflicted remains to be seen.
On a side-note, I mentioned the philosopher Mach above. Bhaskar mentions Mach only in passing, but Skidelsky provides a fine overview of Mach's contributions to the philosophy of science, arguing that he, rather than the better known Comte, has far more impact on the formation of positivistic currents.