This decision concerns an application relating to insurance-risk prediction. However, the EPO refused the application as an obvious implementation of a non-technical scheme. Here are the practical takeaways from the decision T 2626/18 (Insurance risk prediction/SWISS RE) of September 28, 2022, of the Technical Board of Appeal 3.5.01.
The Board in charge summarized the invention underlying the present decision as follows:
2.1 The invention concerns insurance-risk prediction and provides a model analysing potential losses of a company to be insured in order to determine the price of the company’s insurance policy (originally filed application, page 2, line 6 to page 4, line 15).
The model analyses a hypothetical scenario, in which an event causes a loss to the company (page 18, line 1 to page 19, line 9). While not explicitly disclosed, but argued by the appellant during oral proceedings, such an event could be, for example, an accident on the company’s premises. Looking at the Table on page 35 of the original application, the model contains interconnected components called liability risk drivers or LRD members. For example, there is a liability risk driver predicting possible property damage and human injuries resulting from human error (page 44). Another liability risk driver predicts the amount awarded by courts to injured persons as a result of mass litigation (page 35, line 19, to page 36, line 25). The model combines the output of the liability risk drivers and calculates the expected loss cost (page 51, line 26 to page 52, line 8). As shown in the third column of the aforementioned Table, the liability risk drivers employed by the main embodiment analyse business and legal factors only.
3. The claimed invention
The claimed invention additionally assigns to the liability risk drivers physical parameters acquired by measuring devices. The application is not specific as to what sort of physical parameters are used; it discloses merely that the measuring devices “can comprise…all kind of sensors and data capturing or data filtering devices” (page 14, lines 9 to 11). The application does not disclose any embodiment in which particular sensor measurements are processed.
Furthermore, the claimed invention comprises a loss resolving unit that resolves an unspecific loss occurring at a so-called loss unit.
The claims do not provide any technical details of the computer implementation. The application merely states that the claimed units can be implemented in software (page 15, lines 20 to 22).
Fig. 1 of EP 2 461 286 A1
System Claim 1 of Auxiliary Request I
A forecasting and signaling system for automated and automatically tuned operation of a loss resolving unit (40) by means of a control unit controller (10) by forecasting loss frequencies and severities based on captured measure parameters of measuring devices (201,…,261), interacting electronically by signal generation modules and appropriate signal generation, whereas the signal generation is based on forecasted frequencies associated to future loss and loss distributions for individual risks of a plurality of operating units (30) by means of the control unit controller (10), the operating units being exposed to said risk measurable by physical parameters for causing a loss at a loss unit (20,…,26), whereas in case of an occurring loss at a loss unit (20,…,26) the system comprises measuring devices (201,…,261) to scan for, measure and transmit measure parameters to the control unit controller (10), and whereas the control unit controller (10) comprises means to operate the automated loss resolving unit (40) resolving the occurred loss, the loss resolving unit (40) comprising dedicated repair nodes comprising automatic or semiautomatic systems to maintain operation or recover loss of the loss units (2) in case of the occurring loss, characterized
in that the control unit controller (10) comprises a trigger module to dynamically scan, monitor and capture for measuring devices (201,…,261) assigned to the loss units (20,…,26) for measure parameters and to select measurable measure parameters capturing a process dynamic and/or static characteristic of at least one liability risk driver (311-313) by means of the control unit controller (10), each liability risk driver (311-313) representing a measurable real-world liability exposure (31) of an operating unit (30), wherein measuring possibilities at the measuring devices (201,…,261) are dynamically captured and the at least one liability risk driver (311313) is generated based on the on the captured measuring devices (201,…,261) and assigned to the currently measured parameters, and wherein the measuring parameters associated with the generated liability risk drivers (311-313) are measured and transmitted to the control unit controller(10),
in that the control unit controller (10) comprises a driver selector (15) to select a set (16) of liability risk drivers (311-313) parameterizing the liability exposure (31) of the operating unit (30) and dynamically assign the measured measure parameters to the liability risk drivers (311-313), whereas a liability exposure signal of the operating unit (30) is generated based upon measuring the selected measure parameters by means of the measuring devices (201,…,261),
in that the driver selector (15) comprises means to dynamically adapt the set (16) of liability risk drivers (311-313) varying the liability risk drivers (311-313) in relation to the measured liability exposure signal by periodic time response, and adjust the liability risk driven interaction between the loss resolving unit (40) and the operating unit (30) based upon the adapted liability exposure signal,
in that the control unit controller (10) comprises a scenario generator (131) for generating loss scenarios, wherein loss scenarios are given by variables of the control unit controller (10) connecting the liability risk drivers (311-313) to form a function structure, wherein for each scenario a loss model is generated with a frequency distribution function assigned and wherein an expected loss in generated by an aggregator (135) based on a model input and a generated frequency of losses out of the frequency of events and a distribution of a number of losses per event, and
in that a structure of the currently used liability risk divers (311-313) is adapted by the control unit controller (10) by generating and assigning the appropriate liability risk drivers (311-313) based on the currently scanned measure parameters, wherein the liability risk driver structure is based on the generated scenarios, and wherein a specific risk is decomposed by the scenarios into system components of the control unit controller (10), on which the risk drivers (311-313) act independently.
Is it patentable?
In accordance with the first instance examining division, the Board considered that an appropriate starting point to arrive at the claimed invention (according to Auxiliary Request I) is a computer system connected to sensors (rather than just a computer). Then, the Board defined the distinguishing features as follows:
4.5 The claim differs from this starting point by the control unit controller, its sub-units, the loss units and the loss resolving unit.
Afterwards, the Board commented on the main point in dispute, i.e. whether the distinguishing features contribute to the solution of a technical problem, as follows:
4.6 The main point of dispute in this appeal is whether these distinguishing features define a technical solution, as argued by the appellant (see section XI., above), or non-technical matter that could be envisaged by the business person and thus be part of the requirement specification given to the technically skilled person, as considered by the examining division.
4.7 Based on the above understanding of the claimed invention, the Board concludes that the distinguishing features relate per se to an abstract insurance model for predicting future losses and resolving losses that have already occurred. The Board agrees with the examining division that this model constitutes a business method excluded from patentability under Article 52(2)(c) EPC.
Against this finding, the Appellant argued that the claimed model could be automatically executed on a computer, thereby replacing human experts in performing the risk analysis. While the Board accepts that the claimed invention predicts losses in a different way from a human expert, the Board also found that this difference cannot establish technical character:
4.9 The Board also accepts that the claimed model predicts future losses in a different way from a human expert.
However, it is established case law that a comparison with the prior art, for example with what humans did before the invention, is not a suitable basis for establishing technical character of subject-matter excluded from patentability or for distinguishing between technical and non-technical features (see T 1358/09, Reasons, point 5.4).
As a result, the Board in charge disregarded the above-mentioned distinguishing features in accordance with the COMVIK approach and thus found that the claimed subject-matter lacks inventive step. Hence, the appeal was dismissed.
You can read the full decision here: T 2626/18 (Insurance risk prediction/SWISS RE) of September 28, 2022.