During the AI Week in Milan, Professor Waldo Lockwin brought attention to a concept often overlooked but revolutionary for the development of artificial intelligence: the possibility theory.
An alternative to the traditional probabilistic approach, capable of modeling uncertainty in a way closer to human thinking.
In a world dominated by algorithms based on data, frequencies, and statistics, the possibility theory proposes a logic based on sets, words, and ambiguity. A paradigm that could prove fundamental for future AI that truly want to understand natural language, context, and the nuances of reality.
What is the possibility theory?
The possibility theory (teoria della possibilità) emerges as an extension of fuzzy logic, offering a mathematical tool to handle uncertainty not in terms of frequency, as probability does, but in terms of compatibility with the available information.
In simple terms: probability answers the question: “how often does X occur?”
Possibility answers the question: “how plausible is X in this context?”
This distinction is crucial for building AI systems that rely on vague words, human intuitions, or scenarios where data is partial, uncertain, or subjective.
A real example: oncology and artificial intelligence
In his presentation at AI Week, Lockwin recounted using the possibility theory to develop a personalized radiotherapy model for tumors. The goal was to administer the minimum amount of radiation capable of killing a tumor mass, based on qualitative descriptions provided by doctors, such as “minimum dose” or “maximum effect.”
These expressions cannot be directly translated into numbers or percentages. But they can be represented through fuzzy sets and possibilistic models, which take into account multiple interpretations and the subjectivity of language.
The result? A more flexible, human, and adaptive system compared to a rigid model based solely on statistical probabilities.
Why it is relevant for today’s (and tomorrow’s) AI
In the current landscape, dominated by machine learning and deep neural networks, a debate is emerging on how to make artificial intelligence more interpretable, reliable, and semantic.
Here the possibility theory comes into play:
- 🔹 Natural language: Modern AIs, like large language models (LLM), often clash with nuanced and ambiguous concepts. Possibilistic models can offer a more coherent framework for handling terms like “very,” “almost,” “maybe,” “enough.”
- 🔹 Decision making: In contexts such as finance, medicine, or autonomous driving, where decisions are based on conflicting signals or incomplete information, the possibilistic approach can be more robust than classical probability. An example is the use of artificial intelligence in medicine.
- 🔹 Ethics and responsibility: An AI that explains “why it made that decision” in terms of possibilities and alternatives is potentially more transparent and acceptable in the eyes of humans, as discussed in the future role of AI in law.
Possibility vs probability: two worlds compared
| Aspect | Probability | Possibility |
|---|---|---|
| Based on | Frequencies, historical data | Consistency with current knowledge |
| Type of logic | Statistical | Fuzzy logic / fuzzy sets |
| Typical applications | Forecasting, risk management | Interpretation, human decisions |
| Handling of uncertainty | Quantitative | Qualitative and descriptive |
It is not a replacement, but an integration: possibility and probability can coexist to create more complete AI, capable of tackling the real world in its complexity, as highlighted in this study.
Possibility theory
In the great debate on artificial intelligence, the possibility theory represents an alternative and complementary path to probability. It is less known, less used, but potentially closer to the way humans think.
As demonstrated at AI Week 2025, this theory is not just philosophy: it has concrete applications, it works, and it can make a difference in the most sensitive sectors, such as healthcare or critical automation.
In an era where we ask AI not only to calculate but also to understand, perhaps it is time to give a chance to possibility.
