In the past I have explored decisions explaining how otherwise abstract features, such as mathematical method computer program features, can imply technical considerations if they relate to or enable parallel processing and hence can give rise to the all-important technical effect. Indeed, this has been successfully used in prosecution to protect fundamental machine learning innovation. The recent decision T 2910/19 is an example of a, per se non-technical, mathematical method of evaluating insurance claims, which nevertheless was implemented based on technical consideration so that the implementation had a technical effect. This decision provides another useful example of the technical effect of implementing mathematical methods related to parallel processing. There are also some nuggets as to how the Board exercised its discretion in relation to some procedural aspects.
In the first-instance decision, the Examining Division found that the claimed method of evaluating insurance claims did not acquire technical character by using a mathematical method, nor by relating to insurance risks from natural disasters. The first instance thus found that the claimed method corresponded to no more than the implementation of non-technical aspects on a general-purpose parallel processing system such as those disclosed in the cited documents. There being no technical effect beyond the obvious implementation using known technical means, the first instance refused the application for lack of inventive step. The applicant made arguments as to why the particular claimed assignment scheme for assigning insured objects to processors was technical, specifically because it considered which processors had the functions needed for processing the object already in their cache. The Examining Division , however, saw the assignment of insured objects to processors as no more than an administrative choice and so did not buy that there was a technical effect. In the appeal hearing, the applicant was able to convince the Board that the scheme of assigning objects to processors, which was not disclosed in the cited documents, had a technical effect, ultimately leading to the appeal succeeding.
Specifically, the applicant argued, if objects were assigned naively, for example in the order in which they arrived, they would be assigned to processors which do not yet have the required function in their cache, requiring a slow memory access to load the function into cache. This overhead was avoided by the claimed subject matter by assigning objects to processors that already had the necessary function in cache, therefore resulting in a more efficient parallel processing method. The Board recalled that the manner in which a method is implemented on known parallel processing hardware could give rise to a technical effect and found that this was the case here. What remained to be decided, then, was whether, looking to solve the problem of efficient parallel processing claim evaluation, the features in question would be obvious or not. As this question was not considered in the first instance decision, the application was remitted to the first instance for a decision on this point.
Besides providing an illustration of how a parallel implementation of a mathematical method can give rise to a technical effect, the case is remarkable as an example of the Board allowing a late amendment (even after the oral proceedings) and remitting the application back to the first instance for a further decision, possibly based on a further search. While both late amendments and remittance to the first instance are and always have been at the Board’s discretion, the recent change of the Rules of Procedure of the Boards of Appeal have raised the hurdle for both these procedural events and seeing the Board’s reasoning as to the “exceptional circumstances” underlying the exercise of discretion in these cases is always useful.
Here, the provisional opinion of the Board before the oral proceedings followed the first instance reasoning but the applicant was able to convince the Board that the invention was based on technical considerations that were hitherto not considered, even if it remained doubtful to the Board that the claims captured this clearly enough. The Board found that it was credible that the applicant could not get through to the Examining Division with their pertinent submissions. Remarkably, the Board considered this to be an exceptional circumstance that justified the late filing of claim amendments and in the end found that the technical effect was indeed borne out in the amended claims. Since the Board found that the cited documents were not concerned with the technical core of the invention, they stated that they had no choice but to remit the application to the first instance for a further decision, possibly after an additional search.
In summary, the case provides some useful insights on which sort of arguments can work to demonstrate a technical effect inside the computer. Although the claims related to a specific use case of insurance claims, as this non-technical context was not considered for technical effect, the Board’s decision provides an illustration of how arguments for a technical effect inside the computer could work for a broad, core AI invention. In addition, the Board’s view that “not getting through” to the examining division with a technical argument is a reason to allow late amendment and remittal to the first instance, is an interesting data point useful when considering appeal strategy (although it remains of course best practice to make as complete a case as possible as early as possible).