Review Of EPO Software Decisions In 2023 – Fin Tech

Decisions of the EPO Board of Appeal in 2023 can largely be
regarded as a continuation of the trends established in 2022. The
findings of 
Enlarged Board Decision G1/19 from 2021 continue to
dominate, with an emphasis on the whole scope of the claim having a
technical effect and some popular earlier precedents being
overruled. Although we have not collected detailed statistics,
there seems to be increasing trend for Boards to refuse to admit to
requests on appeal, even to the extent of whole appeals being
rejected because no requests are admitted. New requests will only
be admitted on appeal in response to truly new circumstances, such
as unexpected new objections raised by the Board of their own
motion. Below we discuss cases of interest or perhaps general
applicability, highlighting some interesting cases relating to
artificial intelligence and digital therapeutics.

Statistics

With 317 cases by Boards 3.5.01 and 3.5.03 to 3.5.07 in 2023,
there is a clear increase in output compared to 2022 but still no
return to pre-Covid levels and no apparent reduction in pendency
times.

Overall, rejection rates remain high with 70% of cases resulting
in the application or patent in suit being refused or revoked. Of
those that do survive, there has been a slight shift towards
remittal for further prosecution (13%) versus grant or maintenance
(9%). The remaining decisions include cases where the appeal is not
followed through by the applicant, deal with purely procedural
issues, or all requests are rejected as inadmissible. Rejection
rates for these Boards at 70% are consistent with previous years,
as are rejection rates for mixed inventions across all Boards at
88%.

Artificial Intelligence

For a year in which artificial intelligence, in particular large
language models like ChatGPT, has been so prominent in the general
media, there have been remarkably few EPO appeal decisions relating
to inventions involving AI.
T 0183/21 (Controlling the performance of a recommender
system/BRITISH TELECOMMUNICATIONS) of 29-09-2023 perhaps
well illustrates the approach of the EPO to such inventions: in
general terms applying AI to a particular problem is not inventive
and applying AI to a non-technical problem does not in itself
confer technical character, but technical details of the solution
can be inventive. In this specific case, recommending products,
specifically media content, does not have technical character
(following 
T 1869/08 and 
T 0306/10). However, a technical effect to reduce the use of
network bandwidth to provide training data to the recommender
system and the storage necessary for storing training data was
recognised and “achieved, on average, over substantially the
whole scope of the claim”. This effect was achieved as a
result of a trade off with the achievement of a performance metric
that was not suggested in the prior art.

In spite of advancing every conceivable argument, the applicant
in 
T 0761/20 (Automated script grading/UNIVERSITY OF
CAMBRIDGE) of 22-5-2023 was unsuccessful. The invention
related to “a method of automated script grading using machine
learning, which is effectively a computer implemented process. Such
processes may have technical effects – and thus be deemed to
solve a technical problem – at their input or output, but
also by way of their execution (see G 1/19, reasons 85). A
technical effect may also be acknowledged in view of their purpose,
i.e. an (implied) technical use of their output (see G 1/19,
reasons 137).” Most interesting are the discussions of
technical effects “within the computer” and by implied
use.

On the first point, the claimed method contains steps for
extracting numerical “linguistic” vectors from scripts, a
step of training a perceptron, and a step of using the perceptron
to grade the scripts. The extraction of linguistic vectors was not
detailed in the claim and therefore in the eyes of the Board
“cannot be considered to provide any contribution on its own,
be it related to the script acquisition (e.g. scanning or OCR) or
modelling, or to any optimization within the computer.”

The claimed perceptron model is a linear mathematical function
that maps input numerical vectors to output grades and the only
details claimed related to optimization of training to preserve the
ranking of grades, as opposed to minimizing the absolute error in
output grades. According to the Board, “[t]he model is not
based on technical considerations relating to the internal
functioning of a computer (e.g. targeting specific hardware or
satisfying certain computational requirements), and the preference
ranking is chosen merely according to its educational purpose,
which does not relate to any effects within the computer
either.”

On the second point, the applicant argued that the problem
solved by the invention, “providing a computer system that can
automatically grade text scripts [and provide grades] that
correlate well with the grades provided by human markers” is
technical. To decide whether this is technical or not, the Board
considered (i) whether this problem is, or implies, a technical
one, and (ii) whether it is actually solved.

On question (ii), “the Board remarks that the human grading
process is a cognitive task in which the marker evaluates the
content of the script (e.g. language richness and grammatical
correctness) to assign a grade.” They also noted that this
process “is also at least partly subjective: the marker will
have preferences as to style and language, and will be influenced
by experience and grades assigned to scripts in the past.”
Hence, they doubted “that the problem of automating script
grading is defined well enough that one can properly assess whether
it has been solved, i.e. in the sense that it provides a system
that can actually replace different human markers and provide
“correct” grades.

On question (ii), the Board ‘further notes that the field of
“educational technology” as defined by the Appellant …
is a rather inhomogeneous one, covering insights from – and
presumably contributions to – a wide range of
“fields”, technical ones and non-technical ones. It
appears questionable, therefore, that this field can be considered
a technical one as a whole.’

T 0702/20 (Sparsely connected neural
network/MITSUBISHI) of 7-11-2022 discusses neural networks
at some length and in particular the issue of whether an improved
structure of a neural network can provide a technical effect within
a computer. In this case, the difference between the claimed
invention and the prior art was that the different layers of the
neural network are connected in accordance with an error code check
matrix. The applicant asserted that this improved “the
learning capability and efficiency of a machine by reducing the
required computational resources and preventing overfitting”.
The neural network was not claimed in the context of any specific
technical problem. Refusing the application, the Board observed
that the “claim as a whole specifies abstract
computer-implemented mathematical operations on unspecified data,
namely that of defining a class of approximating functions (the
network with its structure), solving a (complex) system of
(non-linear) equations to obtain the parameters of the functions
(the learning of the weights), and using it to compute outputs for
new inputs. Its subject matter cannot be said to solve any
technical problem, and thus it does not go beyond a mathematical
method, in the sense of Article 52(2) EPC, implemented on a
computer.”

The Board’s “Further remarks” suggest that it will
be difficult to convince this Board (3.5.06) at least that a
general invention in the structure or training methods of a neural
network is technical. The Board says that neural networks must
“be sufficiently specified, in particular as regards the
training data and the technical task addressed.” To rely on a
technical effect “within the computer” would likely
require a limitation to specific computer hardware.

Although outside the scope of this paper, it is worth directing
attention to our briefings on two developments in the UK: 
a final determination by the Supreme Court that an
artificial intelligence cannot be an inventor and 
a finding by the High Court (said to be under appeal by
the IPO) that an artificial neural network is not a computer
program as such.

Whole Scope

The greater emphasis on ensuring that an invention meets the
requirements of the EPC across the whole claim scope continues
since G 1/19, even in fairly simple cases. For example in 
T 1887/20 (Input device with load detection and vibration
units/KYOCERA) of 3-3-2023 the appellant argued ‘that
the haptic effect provided by the invention solved the problem
“to provide a realistic sensation of operating a push-button
switch”.’ However, the Board considered this aim not to be
met across the whole scope of the main request, which had no limit
on the duration of the haptic effect and so “encompasses
durations significantly longer than the time a push button is
typically pressed, thus providing feedback even when the button has
been released.” An auxiliary request that did include a limit
on the duration was however considered inventive.

The whole scope requirement is sometimes criticised as
unrealistic since it is almost always possible to find something
covered by a claim to an apparatus or method that does not work
(e.g. a claim to a teapot does not exclude that it is made of
chocolate), therefore the more nuanced approach taken by the Board
in 
T 0814/20 (Adapted Visual Vocabularies/CONDUENT) of
20-3-2023 is welcome. The invention related to image matching and
was supported by a single embodiment directed to vehicle license
plate identification. Initial claims that specified measuring image
“similarity” were considered vague and not serving a
technical purpose. However, claims limited to reidentification of
objects in different images were considered to have a technical
purpose, “because it is tantamount to an objective measurement
in physical reality: is the object observed now the same as the one
observed earlier?”

The remaining issue was therefore whether the claim provided a
technical effect over substantially its whole scope. Having
accepted that the theoretical assumptions underlying the invention
were credible, the Board’s comments are helpfully
pragmatic:

‘The claimed method will not “work” under all
imaginable circumstances. It is probably safe to say that no
computer vision method does. For instance, the present method may
fail to re-identify objects largely changing appearance. However,
the skilled person will understand, from the present claims and the
description, the kind of situations and its parameters (such as
illumination and geometry) for which the method is designed. The
method credibly works over that range of situations.

In the Board’s judgment, this is sufficient to satisfy the
requirement that, in the present case, a technical effect is
present over substantially the whole scope of the claims (see again
G 1/19, reasons 82).’

One approach to an objection that a claim does not solve a
technical problem over its whole scope is to advance a less
demanding problem, or to phrase the problems as to be solved in
certain conditions. However, 
T 1890/20 (Display Device/NEC) of 1-3-2023 makes it clear
that this strategy only works if the claims are limited to the
“certain conditions”. On the other hand, if a claim has
two distinguishing features and one credibly solves a problem
across the whole scope of the claim, it does not matter if the
other distinguishing feature does not solve a problem: 
T 1573/21 (Determining Virtual Machine Drifting/HUAWEI) of
30-08-2023.

Inventive Step

Board 3.2.02, whose caseload normally relates to medical and
veterinary science, applied the Comvik approach in 
T 2165/19 (Taste Testing System/ OPERTECH BIO, INC) of
05-12-2023 and took an interesting approach to the selection of the
starting point for an inventive step objection. The invention
related to “a device aimed at technically implementing a
taste-testing procedure in which a taste sample is presented to a
human subject for tasting and feedback is then gathered from the
subject”. It was noted that such a taste-testing procedure is
not of a technical nature per se (similarly to the odour selection
procedure discussed in 
T 619/02) but what was claimed was a physical device adapted to
automate the method, which is technical. The Board considered that
the Examining Division had incorrectly applied the Comvik approach
based on a document that disclosed automated pipetting systems that
shared some physical features with the claimed device but for a
very different purpose: transferring defined amounts of liquids
between preselected groups of reaction containers.

The Board considered this document not to be an appropriate
starting point for assessing inventive step of claim 1 as the
skilled person would not have looked at this document without the
benefit of hindsight. Instead, the starting point for the invention
should be prior art in the field of devices and methods for
assessing a subject’s response to stimuli. Although the problem
to be solved was considered non-technical and therefore
“given” to the person skilled in the art, it seems
reasonable that it is not obvious to seek a solution to that
problem in hardware for a different purpose. At the same time, this
is consistent with many cases where general
purpose hardware is considered a suitable starting point
for implementation of non-technical methods.

General purpose hardware, such as computers and networks, are
often considered “notorious”, meaning that no specific
prior art disclosure need be cited.
T 1898/20 (Method and server for providing air fare
availabilities/SKYSCANNER) of 05-12-2023 warns that care
must be taken in asserting that something is notorious. The
invention here related to assembling data relating to air fares and
seat availability. The claims referred to a “distribution
system server” which implements specific functions. Although
the distribution system server was discussed in the prior art
section of the application, the applicant argued that these
mentions were not necessarily admissions of common general
knowledge. The Board noted that, in contrast to US Patent Law, the
EPC does not know the principle of admitted prior art and so could
not assume the distribution system server is notorious. Therefore
the case was remitted to the examining division for further
prosecution, in particular to carry out a search for a prior art
document disclosing the distribution system server.

There was a similar outcome in 
T 2321/19 (Capturing user inputs in electronic
forms/BLACKBERRY) of 13-2-2023 where the Board agreed with
the applicant’s argument that it was very difficult, thirteen
years after the date of filing of the present application, to
assess what was the common general knowledge of the person skilled
in wireless hand-held devices at that date of filing, especially
since the technology of mobile phones had evolved very quickly at
that time. Since this aspect of the common general knowledge of the
skilled person was highly relevant, the case was remitted to the
examining division to allow for two-instance consideration of the
common general knowledge.

The scope of notorious prior art and common general knowledge
was also at issue in 
T 1273/20 (Performance storage system/EMC) of 13-11-2023.
The Board observed that ‘no specific documentary evidence may
be needed to prove knowledge which belongs to the “mental
furniture” of the skilled person, such as routine design
skills and general principles of system design which are often
necessary just to understand the prior art in the relevant field (T
190/03, Reasons 16).’ And went on to conclude that
“[m]emory hierarchies are so pervasive in the computing field
that the board considers that no documentary evidence of them is
needed.” Contrasting with the two cases discussed above, it
was only the general concept of memory hierarchies that was
considered common general knowledge and sufficient to render the
claimed invention obvious, and not any detailed implementation
thereof.

The absolute novelty approach of the EPC implies that all prior
art disclosures are of equal potential as starting points for an
inventive step argument. In 
T 1092/19 of 04-10-2023 the Board rejected an argument
that the person skilled in the art would not consider modifications
to a method described in a working draft of a video coding standard
because of the nature of that document, rather than based on
technical reasons. The Board commented “the person skilled in
the art is motivated by the desire for further improvement and is
not dissuaded from their pursuit by administrative decisions, e.g.
those taken by standardisation organisations.”

That an invention is a straightforward automation of a known
manual method is a fairly common reason for asserting a lack of
inventive step. However, 
T 0302/19 (Cell characterization/BIO-RAD) of 21-12-2023
cautions that “[f]or such an argument to succeed, it should be
clear what is the alleged manual practice, it should be convincing
that it was indeed an existing practice at the relevant date and
that it would have been obvious to consider automating it.” In
that case, the detail was lacking and the alleged manual procedure
unconvincing as it would have been too laborious to carry out
manually.

J A Kemp LLP acts for clients in the USA, Europe and
globally, advising on UK and European patent practice and
representing them before the European Patent Office, UKIPO and
Unified Patent Court. We have in-depth expertise in a wide range of
technologies, including
Biotech and Life Sciences,
Pharmaceuticals,
Software and IT,
Chemistry,
Electronics and Engineering and many others. See our
website to find out more.

The content of this article is intended to provide a general
guide to the subject matter. Specialist advice should be sought
about your specific circumstances.