A review of BL O/0894/26 Emotional Perception AI Ltd: Can a Subjective Improvement Ever Be Technical?

The refusal of BL O/0894/26 (Emotional Perception AI Ltd) based on lack of inventive step was, from a European perspective, hardly surprising. After the case was remitted back to the examining division, most practitioners were expecting the applicant to struggle with demonstrating technical character and hence inventive step. Nevertheless, I found the reasoning in the decision to be largely unsatisfactory overall. 

In particular, the refusal presupposes that a “semantically better” file recommendation is “not enough to be considered a technical effect” because it is “better only in a subjective psychological sense” (para. 53). In this article, I challenge this line of reasoning, which I think, at best, is incomplete. 

Before I begin, I should add that my challenging of this line of reasoning is not based on any European case law (to the contrary, there is European case law to suggest that the conclusion is valid) but rather a broader reflection of whether I think the conclusion ought to be justifiable.  

The simplest way to challenge the reasoning is to present a counterexample. Consider a new and inventive manufacturing process for producing a textile having a visually aesthetic texture. The impression of the textile (e.g., its “feel” and “look”) onto the user is fundamentally subjective. Nevertheless, most practitioners would agree that the textile and its method of manufacture are patentable. That is, inventions that derive value from the subjective impression they confer on a user are not inherently devoid of technical character. On this basis, the hearing officer’s line of reasoning in paragraph 53 seems incomplete.

It seems to me that it is also necessary to consider if the output of the method is technical in and of itself and whether the method steps for determining the output are technical, irrespective of whether the output is deemed technical. These elements are discussed in the decision, but they are not appropriately distinguished. This has led the hearing officer to conclude a lack of technical effect simply on the basis that the file merely presents a subjectively better result. To an extent, it is also relevant whether the properties of the output are a direct consequence (i.e., emerge) from the method steps, but I will not consider this in any more detail in this article.

Regarding the first consideration, I see no reason why a media file in and of itself ought to be regarded as non-technical. While a media file may encode contents that are non-technical (e.g., a poem or a business method), this should not detract from the fact that it physically exists and is objectively identifiable. As it contains “audio data”, “video data” or “static images”, it also does not fall squarely within any one of the categories defined as excluded subject matter under Section 1(2) UKPA. I note that the hearing officer does seem to essentially agree with this position, although this did not appear to fully feed into their inventive step analysis. 

Regarding the second consideration, the situation is less positive. While the manufacturing steps of the textile are plainly technical, the steps of the claimed method essentially relate either to the training of an ANN, which may be regarded as a mere mathematical method, or to using the trained ANN to assist with selecting a media file with semantic similarity to target data.

To assess whether these method steps contribute to technical character (noting that non-technical features can contribute to technical character), it seems necessary to consider whether the claimed steps inherently map onto a technical aspect of the file (rather than some arbitrary characteristic of it).

In the context of this case, this appears to boil down to the following:

Is, identifying measurable characteristics of a media file that correlate with users perceiving those files as being semantically similar, technical?  

Is, selecting a media file that has determined measurable characteristics that correlate with users perceiving those files as being semantically similar to target data, technical?  

I am not convinced that either statement can obviously be dismissed as non-technical. A user’s perception is subjective, but it can nevertheless be studied through objectively measurable indicators and, across a population, the effectiveness of the method could be tested using known statistical methods (in the same way that the effectiveness of a new drug may be tested in different sub-populations and be found to have differing efficacy across those sub-populations). That is, the fact that a predicted outcome is subjective does not necessarily imply that the predicted subjective outcome cannot be objectively measurable.

There is a separate evidential issue of whether it is credible that these effects are achieved (not all people think or perceive the same) and in this regard, the application does fall short. But none of this was, in my opinion, fully considered by the hearing officer in his decision.

Subscribe to receive more articles like this here.