Experts are not always experts in everything, and they are not always free from bias. That’s why, when engaging in foresight activities—predicting technological scenarios using expert panels—one must be cautious of the Unknowns. And for experts, admitting unfamiliarity with a subject, especially when they’ve been brought in for their knowledge, isn’t easy. Yet the true value often lies in these “Unknown Unknowns” (Rumsfeld, 2002).
In this article, we explore a promising technology that experts failed to identify—what we might call a false negative, something initially interpreted as an error in analysis. False negatives are often the inevitable result of overwhelming amounts of information. Experts are deeply knowledgeable about the technologies within their direct field of control and monitoring. They can keep up with the literature, attend conferences, or talk with peers—but only within a limited number of related technologies. They cannot monitor all developments. Fortunately, patent analysis can significantly support us in this regard.
When Algorithms Help Experts
In a 2010 technology foresight project for a client in the biomedical sector, we examined 6,107 patents after identifying trends in Laser and Ultrasound technologies. These patents were grouped into four categories (Group A61B 8/00: Diagnostic Ultrasound; Subgroup A61B 18/18: Laser Therapy; Group A61N7/00: Ultrasound Therapy; Subgroup A61B17/02: Devices to Keep Wounds Open). All patents dated from 2000 to 2010—a massive amount of information. The challenge was to extract insights from such a large dataset without asking experts to read and comment on every patent—avoiding the need for extensive domain knowledge to interpret every detail—and instead use experts’ knowledge to confirm or challenge the results of the large-scale analysis.
We used a functional approach based on reconstructing the causal chains that lead a product to function, rooted in German-style Engineering Design. In this specific case, the approach also allowed us to identify emerging technologies and conduct predictive studies. The strength of this approach lies in combining powerful semantic analysis technologies for automated data processing with interpretive capabilities typical of functional product analysis.
The Discovery of “Bone Age Technology”
A remarkable result is shown in the table below. The algorithm was able to identify a technology—labeled “bone age technology”—and to extract the key functional verbs and closely associated terms, allowing even non-experts to accurately interpret its content.
Table: An example of an emerging technology in the field of ultrasound devices.
The process clearly highlighted the expression “bone age” as a new and significant term in the patent database. At first, it wasn’t obvious what technological innovation was involved, and the experts had not mentioned this technology in previous interviews (were they unaware of it? Or had they simply forgotten to mention it?).
Discover our AI software for the automatic extraction of information from technical documents.
So, Did the Algorithm Get It Wrong?
The analysis was repeated a couple of times because the experts believed the algorithm had generated an artifact—“Everyone knows ultrasound is used for soft tissue.” However, upon repeating the analysis, the result remained unchanged, and in fact, the algorithm was able to pinpoint this emerging technology in great detail.
By arranging the words according to their syntactic/semantic role in a possible sentence, we could provide the following functional definition: “Bone age” [relates to the] “evaluation/measurement” [of] “skeletal/metaphyseal/wrist” “growth” [in] “children” [or] “premature infants,” [or in] “osteoporosis” [in] “adults,” [using] “ultrasound scanning.” No mistake here.
So what went wrong in the experts’ interpretation? Nothing, except their perspective. The experts correctly believed that bones couldn’t be easily visualized with ultrasound—but they hadn’t considered pores, decalcifications, cartilage, etc.—all “transparent” to ultrasound!
Once the mystery was explained, the experts confirmed that “bone age” refers to the degree of bone maturation in a child. It is especially important to assess bone age in premature infants by examining the wrist bones. Traditional bone scans use X-rays, which can be harmful to newborns. One of the most promising research directions is, in fact, using ultrasound, which is much safer. Back in 2010, only one device based on this technology was available. The same applies to osteoporosis in elderly patients or other diseases—also detectable via ultrasound, and far less harmful than a radiograph.
In short, while experts failed to identify this opportunity, the algorithm—trained functionally and applied to patent data—was able to detect it, even though only a small number of patents described it, buried among thousands of other applications.
Innovation Hides Where You Least Expect It
The story of “bone age technology” teaches us a fundamental lesson: technological innovation does not always follow expected paths, and sometimes it hides in plain sight—right where no one is looking. In this context, the combined use of semantic algorithms and functional analysis has shown great potential. It’s not about replacing human intuition or expert experience, but about amplifying their reach—helping uncover weak signals and hidden connections. Because in the end, the real challenge is not predicting the future, but being ready to recognize it when it shows up in unexpected forms.
Request a demo of our AI tools.
