If so, the credibility of an invention ultimately comes down to one question:
Does the invention really produce the claimed effect?
There are two main ways to explain the effect of an invention:
・Evidence-based inventions
・Logic-based inventions
An evidence-based invention follows the principle of "proof over theory." The existence of the effect—and therefore the validity of the invention—is supported by evidence such as experimental data.
A logic-based invention explains the causal relationship between the invention's features and the resulting effect through a step-by-step chain of reasoning.
Evidence-Based Inventions
Even if the mechanism behind an effect is unknown, an invention can still be valid if there is evidence showing that a specific configuration produces a specific effect.
For example, suppose a medicine has a beneficial effect. Even if we do not understand why the medicine works, the medicine can still be considered an invention.
In some cases, such as psychotropic drugs, the discovery of the drug comes first, and only later do researchers understand how the human body reacts to it.
For this type of discovery-based invention, evidence demonstrating the effect can establish the credibility of the invention, even when the mechanism remains unknown.
Another example involves antenna design.
Suppose engineers design an antenna element with a particular electrode shape and evaluate its performance using simulation software. The simulation shows improved radiation characteristics.
A theoretical explanation based on electromagnetic engineering may eventually be possible. However, when the simulation data is first obtained, the exact reason for the improvement may still be unclear.
Because the simulation data supports the claimed effect, this type of invention is also evidence-based.
In other words, it is possible to obtain a patent even if the mechanism of the invention cannot be fully explained.
Challenges in Patent Examination
Patent examination of evidence-based inventions is actually very difficult.
If an applicant submits incorrect or even fraudulent data, it is extremely difficult for a patent examiner to detect the problem.
Patent attorneys also generally prepare patent specifications based on the assumption that the data provided by the applicant is accurate.
Of course, if misconduct or errors in the data are later discovered, the basis for the invention may disappear, and the patent can be invalidated.
How Much Improvement Is Enough?
Evidence-based inventions also raise an important question:
How should performance improvements be interpreted?
If an invention doubles performance, most people would consider it a major technological advance.
However, what if performance improves by only 2%?
Is a 2% improvement significant enough to deserve patent protection?
A 2% difference might simply fall within the margin of error. On the other hand, achieving even a 2% improvement could be extremely difficult and highly innovative, depending on the technical field.
Logic-Based Inventions
Some inventions allow the mechanism that produces a particular effect to be explained logically.
Many software patents and electronic circuit inventions fall into this category.
Consider a vehicle that warns a driver when a collision with a pedestrian is likely.
The vehicle may:
1. Capture video using an onboard camera.
2. Detect pedestrians through image recognition.
3. Determine the pedestrian's position and direction of movement.
4. Calculate the distance between the vehicle and the pedestrian.
5. Analyze the vehicle's speed and travel direction.
6. Predict whether a collision course exists.
7. Estimate the remaining time before collision.
8. Issue an audio warning if the remaining time falls below a threshold.
Through this logical sequence, it can be explained that the invention reduces the risk of collisions between vehicles and pedestrians.
If the logical structure is sound, most people can understand that the invention works even without experimental proof.
Inventions Created Through Imagination
Logic-based inventions can often be created through imagination alone.
However, every step of the logic must be carefully built and connected.
If the logic contains weaknesses or inconsistencies, those problems may later be exposed during patent enforcement, even if the patent examiner overlooked them and granted the patent.
Logic Supporting Evidence
Even evidence-based inventions can be strengthened by logical explanations.
Consider an invention in which the brightness of a display device is increased by adding certain chemical compounds.
If the mechanism is unknown, the invention is primarily evidence-based.
In many patent applications of this type, the inventors simply present experimental conditions and results as proof that the invention works.
However, in one application I reviewed, the inventors did more than present experimental data. They also proposed a hypothesis explaining why the added compound increased brightness.
Because it was only a hypothesis, there was no guarantee that it was correct.
Nevertheless, combining experimental evidence with a plausible explanation made the invention much more convincing.
Evidence Supporting Logic
Logic-based inventions can also be strengthened through evidence.
For example, inventions related to natural language processing (NLP) can logically explain how a new algorithm interprets language.
However, the credibility of the invention becomes much stronger if performance metrics such as precision, recall, or accuracy are also presented and shown to improve.
In one case involving an electronic circuit invention, we wanted a U.S. patent examiner to appreciate how remarkable the invention was.
To support the patent application, we obtained a declaration from a well-known university professor. The professor explained why the inventive concept was highly impressive from the viewpoint of an expert in the field.
Opinions from independent experts can also serve as a form of evidence.
How Artificial Intelligence Is Changing Software Inventions
Software is fundamentally an invention of logic.
If a software process is logically possible, it should also be physically implementable. Therefore, software inventions can normally be explained through logical reasoning.
Traditionally, there was nothing mysterious about software.
However, the rapid development of artificial intelligence technologies, especially deep learning, has begun to change this assumption.
AI systems learn independently and generate their own internal decision-making criteria.
As a result, humans are increasingly unable to understand exactly how an AI reaches a particular conclusion.
The Example of AI Shogi Programs
Some AI-powered shogi programs (shogi is a traditional Japanese board game often compared to chess) analyze enormous collections of game records to learn strategic thinking from expert players.
They further improve by repeatedly playing against themselves.
After enough self-training, humans can no longer fully understand the strategies the AI uses when making moves.
The Example of Hit Song Prediction
There are also AI systems designed to predict whether a new song will become a hit.
These systems learn from massive collections of music and identify patterns associated with successful songs.
However, developers often cannot clearly explain which aspects of the music the AI is evaluating when it predicts success or failure.
AI Patents and the Black Box Problem
AI-related inventions differ fundamentally from traditional software inventions because of the black-box nature of their algorithms.
For this reason, some AI inventions may need to be treated more like evidence-based inventions than logic-based inventions.
There is also an active research field known as Explainable AI (XAI), which aims to make AI decision-making processes understandable to humans.
At present, it remains unclear how patent practice for AI inventions will evolve.
At the same time, technological fields with few precedents and uncertain futures often provide excellent opportunities for obtaining strong patent protection ahead of competitors.
For innovators, uncertainty is not only a challenge—it is also an opportunity.