Object-centric proto-symbolic behavioural reasoning from pixels
Clinical Snapshot
PICO Framework
| P — Population | Synthetic 2D and 3D environments (dSprites variants) |
| I — Intervention | Novel brain-inspired deep-learning architecture using object-centric representations |
| C — Comparator | Not explicitly stated - appears to be baseline performance without object-centric reasoning |
| O — Outcomes | Emergent conditional behavioural reasoning, logical composition, XOR operations, environmental control, adaptation to changes |
Bottom Line
This computational study presents a novel deep-learning architecture that learns object-centric representations to bridge sensory input and abstract reasoning. While the approach shows promise for emergent logical reasoning and environmental adaptation, the work is limited to synthetic 2D/3D environments that fall short of real-world complexity. The study lacks statistical validation, comparison with existing methods, and clinical applicability. For healthcare professionals, this represents early-stage AI methodology research with potential future implications for medical decision support systems, but no immediate clinical relevance. The work contributes to the theoretical foundation of AI systems that might eventually support clinical reasoning, but significant development would be needed before any healthcare applications could be considered.
Key Findings
Effect Size: Not quantified
Primary Outcome: Successful demonstration of emergent conditional behavioural reasoning
Nnt Or Sensitivity: Not applicable - proof-of-concept study
Confidence Interval: Not reported
Clinical Application
Currently limited to research settings with synthetic data No direct relevance to Australian clinical practice, TGA, or PBS. May have future implications for medical AI development Not applicable to clinical populations
Abstract
Autonomous intelligent agents must bridge computational challenges at disparate levels of abstraction, from the low-level spaces of sensory input and motor commands to the high-level domain of abstract reasoning and planning. A key question in designing such agents is how best to instantiate the representational space that will interface between these two levels-ideally without requiring supervision in the form of expensive data annotations. These objectives can be efficiently achieved by representing the world in terms of objects (grounded in perception and action). In this work, we present a novel, brain-inspired, deep-learning architecture that learns from pixels to interpret, control, and reason about its environment, using object-centric representations. We show the utility of our approach through tasks in synthetic environments that require a combination of (high-level) logical reasoning and (low-level) continuous control. Results show that the agent can learn emergent conditional behavioural reasoning, such as (A → B)∧(¬A → C), as well as logical composition (A → B)∧(A → C)⊢A → (B∧C) and XOR operations, and successfully controls its environment to satisfy objectives deduced from these logical rules. The agent can adapt online to unexpected changes in its environment and is robust to mild violations of its world model, thanks to dynamic internal desired goal generation. While the present results are limited to synthetic settings (2D and 3D activated versions of dSprites), which fall short of real-world levels of complexity, the proposed architecture shows how to manipulate grounded object representations, as a key inductive bias for unsupervised learning, to enable behavioral reasoning.
References
- 1.van Bergen, R., Hübotter, J., Lago, A., & Lanillos, P. (2026). Object-centric proto-symbolic behavioural reasoning from pixels. Neural Networks, 108407. https://doi.org/10.1016/j.neunet.2025.108407
Related Research
Philosophical transactions. Series A, Mathematical, physical, and engineering sciences
A sentence is worth a thousand pictures: can large language models understand hum4n L4ngu4ge and the W0rld behind W0rds?
15 May 2026
Philosophical transactions. Series A, Mathematical, physical, and engineering sciences
What physics offers for artificial intelligence? Lessons from the brain's inner time and its dynamics.
15 May 2026
Neural networks : the official journal of the International Neural Network Society
Semi-supervised classification and projection with adaptive flexible structure optimal graph
9 May 2026
This content is for educational purposes for healthcare professionals only and does not constitute clinical advice. Clinical decisions should be based on individual patient assessment, current guidelines, and appropriate specialist consultation. Editorial Standards · Privacy Policy · Terms of Service