marcel blattner | May, 2025
Image created by marcel blattner. 1Mio. Eigenvalues of a Bohemian Matrix with base pattern: base_pattern=[-1j,0,1]
Agency in biology and computer science
The concept of agency has evolved quite differently in biology and computer science, creating significant challenges for fields like artificial intelligence and synthetic biology. Biological agency emerges as a rich, multi-layered phenomenon involving goal-directedness, autonomy, and meaning-making. However, in computer science, „agency“ often reduces to simply being able to act in response to environmental inputs – a much thinner conception that misses crucial aspects of biological agency.
In biology, as Ball, Levin and others describe, even single cells display remarkable agency through their ability to:
- Maintain their identity while actively exchanging matter with the environment
- Generate their own goals rather than just following programmed rules
- Create internal representations of their environment
- Assign meaning to signals based on their goals
- Act with genuine autonomy rather than pure stimulus-response
- Participate in collective agency at higher organizational levels
In contrast, computer science often treats agents as entities that can:
- Perceive their environment through sensors
- Process inputs according to programmed rules
- Take actions that affect their environment
- Optimize for pre-defined objectives
The missing link
The critical missing elements in the computational view include the self-generated nature of goals, the creation of meaning, and true autonomy. While an AI system may appear to pursue goals, these are ultimately imposed by human designers rather than emerging from the system’s own organization and needs. The system doesn’t truly care about its objectives or create its own meaning from environmental signals. This divergence creates several problems.
The resolution may lie in developing richer theoretical frameworks that can bridge this gap. Some promising directions include:
- Using thermodynamic and information-theoretic approaches to understand how genuine agency emerges from physical principles, as discussed in Ball’s papers and books.
- Developing formal models of goal-directedness and meaning-making that could apply across biological and artificial systems.
- Exploring how collective agency emerges in both biological and computational systems, potentially revealing common principles.
- Studying minimal forms of biological agency (like in bacteria) to understand the essential requirements for genuine autonomous goal-directedness.
Way forward
Progress will require deeper dialogue between biology and computer science. Biologists can help identify crucial aspects of agency that computational approaches are missing, while computer scientists can help formalize and operationalize biological concepts.
The ultimate goal should be frameworks that can deal equally well with biological cells, animal behavior, human cognition, and artificial agents – while recognizing both commonalities and important differences in how agency manifests across these cases.
This conceptual work is not merely academic – it has practical implications for fields from synthetic biology to artificial intelligence. Only by understanding the full richness of biological agency can we hope to create artificial systems that approach the remarkable capabilities of living things.
The fact that such a fundamental concept can be interpreted so differently in two major scientific fields suggests the need for more cross-disciplinary work to align our theoretical foundations. Bridging the agency gap may be essential for advancing both biological and computational sciences in the next decade.
Source
Ball, P. (2023). How Life Works: A User’s Guide to the New Biology. In How Life Works. University of Chicago Press.
Levin, M. (2019). The computational boundary of a “self”: developmental bioelectricity drives multicellularity and scale-free cognition. Frontiers in psychology, 10, 2688
Rosslenbroich, B., Kümmell, S., & Bembé, B. (2024). Agency as an Inherent Property of Living Organisms. Biological Theory, 1–13.
Seth, A. (2024). Conscious artificial intelligence and biological naturalism. PsyArXiv preprint.