Active inference and artificial reasoning
Karl Friston, Lancelot Da Costa, Alexander Tschantz, Conor Heins, Christopher L Buckley, Tim Verbelen, Thomas Parr · 2026
Read Active inference and artificial reasoning by Karl Friston, Lancelot Da Costa, Alexander Tschantz, Conor Heins, Christopher L Buckley, Tim Verbelen, Thomas Parr (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-13601584.txt.
Description
Karl Friston and co-authors wrote a 2026 paper on active inference as a frame for machine reasoning. The core move is to pick actions whose outcomes teach the most about the hidden world model. That rule gives a principled path to learning structure and finding rules.
In this frame, each plan is scored by expected free energy, a mix of expected information gain and value. Model scores can be updated fast from past beliefs, so the gain of each act can be weighed before choice. The authors show the method on a classic three ball task used to model insight. The demo stresses sample efficiency: seek the outcome that cuts the most doubt about the world.
Catalog record
| CREATOR | Karl Friston, Lancelot Da Costa, Alexander Tschantz, Conor Heins, Christopher L Buckley, Tim Verbelen, Thomas Parr |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 92 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |