By Michael Borella --
On June 11, 2019, the U.S. Patent and Trademark Office held a public presentation -- a patent quality chat -- regarding the interpretation of computer-implemented claims using functional language under 35 U.S.C. § 112. Mostly, this presentation was an overview of the USPTO's January 7, 2019 Federal Register notice on the same topic. But, there is a growing likelihood that Congress will strengthen § 112 in an attempt to disambiguate § 101 (see the recent Senate Subcommittee on Intellectual Property testimony and the current draft bill). While most of the presentation was a reiteration of existing law and USPTO practice, taking another look at the current state of § 112 and where it may be going for computer-implemented inventions is a worthwhile effort.
The presentation begins with an acknowledgement that the Federal Circuit has raised issues with broad, functional claiming of software technologies without adequate support in the context of both § 112 and § 101. The USPTO's position is that these problems can be addressed by properly applying § 112.
Claim Construction during Prosecution
To identify claim limitations that invoke § 112(f), examiners are required to apply the three-prong analysis set forth in MPEP § 2181(I):
(A) the claim limitation uses the term "means" or "step" or a term used as a substitute for "means" that is a generic placeholder (also called a nonce term or a non-structural term having no specific structural meaning) for performing the claimed function;
(B) the term "means" or "step" or the generic placeholder is modified by functional language, typically, but not always linked by the transition word "for" (e.g., "means for") or another linking word or phrase, such as "configured to" or "so that"; and
(C) the term "means" or "step" or the generic placeholder is not modified by sufficient structure, material, or acts for performing the claimed function.
When a limitation uses the term "means" coupled with functional language, there is a rebuttable presumption that § 112(f) applies. This presumption can be overcome if the claim also includes structure necessary to perform the functional element. Conversely, when a limitation does not use the term "means" there is a rebuttable presumption that § 112(f) does not apply. In Williamson v. Citrix Online, LLC, the Federal Circuit held that this presumption is overcome when "the claim term fails to recite sufficiently definite structure or else recites function without reciting sufficient structure for performing that function."
Functional limitations that do not use the term "means" often have a generic placeholders instead, such as "mechanism for," "module for," "device for," "unit for," "component for," "element for," "member for," "apparatus for," "machine for," or "system for." The USPTO makes it clear that there is no fixed list of these generic placeholders and that each limitation is to be evaluated on the basis of its own language.
Broadest Reasonable Interpretation of a Limitation that Invokes § 112(f)
Once an examiner determines that a limitation invokes § 112(f), the examiner is to further determine the broadest reasonable interpretation (BRI) of the limitation. According to the presentation, this is "the structure, material, or act described in the specification as performing the entire claimed function and equivalents to the disclosed structure, material, or act." Thus, the examiner must look to the specification in order to properly determine the BRI.
Computer-Implemented Limitations under § 112(f)
Many software claims are functional in nature with these functions being carried out by a general purpose processor or generic computer hardware. This structure is not enough to meet the § 112(f) requirements. Instead, the specification must disclose an algorithm for performing the functional feature. In accordance with MPEP § 2181(II)(B), an algorithm is typically "a finite sequence of steps for solving a logical or mathematical problem or performing a task," and the applicant "may express the algorithm in any understandable terms including as a mathematical formula, in prose, in a flow chart, or in any other manner that provides sufficient structure." In other words, the applicant must avoid purely functional claiming without the specification explaining how the hardware carries out the functional limitations.
If no algorithm is disclosed or the disclosed algorithm is insufficient to perform the claimed function, then the claim is indefinite under § 112(b) and lacks written description (and possibly enablement) under § 112(a). Notably, whether a disclosed algorithm meets this requirement is based on "what one of ordinary skill in the art would understand as sufficient to define the structure and make the boundaries of the claim understandable." Further, one cannot avoid disclosing an algorithm by arguing that a person of such ordinary skill would be able to implement an algorithm to perform the functional limitation.
Written Description and Enablement of Computer-Implemented Inventions
Regardless of whether § 112(f) is invoked, examiners must still carry out a separate § 112(a) evaluation. In this inquiry, the examiner determines whether the specification includes a sufficient written description of the claimed invention. Notably, the specification must include more than just a result achieved by the claims. How much and what detail is required will vary from technology to technology and the level of detail in the claims. For instance, "[i]nformation that is well known in the art need not be described in detail in the specification." Moreover, if a genus is claimed, "disclosure of the species must be sufficient to convey to one skilled in the art that the inventor possessed the subject matter of the genus." But the level and extent of this disclosure will also vary from technology to technology as well as with the breadth of the genus. For results-oriented computer-implemented limitations, this means that the specification should describe how these limitations are achieved and not just what is achieved.
Regarding enablement, the specification must teach one of skill in the art how to practice the entirety of the claimed invention without undue experimentation. When the specification does not support the full scope of a claim, the claim lacks enablement. Thus, the test from In re Wands still applies, and the specification does not have to disclose features well known in the art. The presentation indicates that "[t]his is of particular importance with respect to computer-implemented inventions due to the high level of skill in the art and the similarly high level of predictability in generating programs to achieve an intended result without undue experimentation."
Analysis
It is likely that Congress will try to clear at least part of the § 101 morass by heightening the § 112(f) and § 112(a) provisions regarding the level of disclosure required for software claims. Thus, applicants should be wary of the potential pitfalls to come.
Of course, some would argue that functional claiming should never be used, because claims invoking § 112(f) only cover the corresponding structure, material, or acts described in the specification and equivalents thereof. But, functional claiming is virtually impossible to avoid when protecting certain types of computer-implemented technologies, especially under Williamson and the draft bill. Therefore, any functional claim element should be accompanied by at least a brief description of an algorithm that can be used to carry out the function. Even if the function is basic (e.g., receiving data, sorting a list of numbers, or displaying an image on a graphical user interface) a short discussion of at least one way (preferable two or three ways) of achieving these goals should be included in the specification. If the invention includes an unconventional way of practicing the limitation, this should be described in more detail (and claimed as well).
Results-oriented claims, on the other hand, should be avoided when possible. For instance, the limitation "training a neural network to recognized handwritten numbers" is results-oriented, even if multiple algorithms for doing so are known. Outcomes can vary based on the arrangement of the neural network, the type of training data used, and how the training takes place. Therefore, to provide a reasonably definite claim, the applicant should address each of these issues. Here, I am not suggesting that such a limitation is indefinite without this degree of disclosure, but instead recommending that one err on the side of caution with results-oriented claims. In contrast, the limitation "sorting a list of words so that they are in alphabetical order" could probably be accompanied by no more than a mention, in the specification, of a few well-known sorting algorithms.
One thing that the applicant does not have much control over is how an examiner or a judge will evaluate the level of ordinary skill in the art. Even if computer-implemented inventions may effectively require a high level of skill for the ordinary practitioner, this is ultimately a subjective evaluation. Again, it is advisable to add at least some supplementary description of each claim limitation even if that limitation is well-known in the art. In this way, such disclosure can be a fallback position in the event that the examiner or judge finds that the level of skill in the art is lower than one might expect.
"training a neural network to recognized handwritten numbers""
As broad as this claim is, in practice it would probably be even broader to cover something like:
"1. training a model to recognize a parameter of an input"
2. wherein the parameter is a corresponding number, and the input is digital copy of a human written paper" or something
Software drafters have brought this on themselves.
Also, I may be in the minority, but I dont see the 112(f) draft changes meaning much.Even without touching the law, the Federal Circuit is already going in that direction pretty clearly. Not sure that the draft changes actually change anything at this point.
Posted by: JPosner | June 19, 2019 at 06:48 AM
"[T]he Federal Circuit has raised issues with broad, functional claiming... without adequate support in the context of both § 112 and § 101. The USPTO's position is that these problems can be addressed by properly applying § 112."
The USPTO is quite correct. The overwhelming majority of that which is presently treated as a §101 issue under Alice would be more cogently treated as a §112 issue under Ariad.
Posted by: Greg DeLassus | June 19, 2019 at 11:37 AM
"Even without touching the law, the Federal Circuit is already going in that direction pretty clearly. Not sure that the draft changes actually change anything at this point."
I do not really disagree with you here, but I think that there is significance in the Congress signalling to the CAFC "we agree with what you are doing here." It is not nothing for the CAFC to receive positive feedback from the legislature. At the very least, it locks in the current trend, so that a new en banc majority does not take back what the Williamson en banc majority just gave.
Posted by: Greg DeLassus | June 19, 2019 at 12:08 PM