The Agency Problem: Why Biology Needs to Stop Ignoring Purpose

Biological agency, or the capacity for organisms to act with purpose, is often dismissed as unscientific or relegated to psychology. Yet, as Philip Ball argues, this dismissal creates a blind spot in our understanding of life. By treating organisms as mere machines, we ignore the reality that living things do not just respond to stimuli; they improvise. This conversation shows that the hardwired view of evolution cannot explain how organisms navigate novel environments. For researchers, technologists, and systems thinkers, recognizing agency is not about embracing mysticism. It is about developing a rigorous, mechanistic theory of how complex systems integrate history and context to set their own goals. Those who master this framework will gain an advantage in understanding both biological evolution and the next generation of autonomous AI.

The Failure of the Hardwired Fallacy

Conventional biology often leans on a reductionist view: organisms are complex machines, and their behaviors are scripted by genetic instructions shaped by evolution. Philip Ball challenges this by pointing out a flaw in the logic: evolution cannot hardwire a solution for every specific, novel circumstance an organism encounters.

The ancestors of those organisms have not faced that situation before so there is no way evolution can tell it what to do in every circumstance. It is a far better strategy for evolution to create agents and I think that is what it does.

-- Philip Ball

When we assume every behavior is a pre-programmed script, we miss the systemic reality of behavioral plasticity. An organism does not just execute code; it integrates environmental signals with its own internal history to make contextual decisions. The machine metaphor fails because it implies a predictable, clockwork output. Living systems, by contrast, are fundamentally contingent. By shifting the focus from what is life to what has agency, we move from an unanswerable philosophical debate to a testable scientific framework.

The Hidden Costs of Ignoring Goal-Directedness

Biology has long been uncomfortable with teleology, or the idea that systems have goals. As Ball notes, this discomfort leads to a cryptic reintroduction of agency, where purpose is smuggled back into scientific papers under different names because it is impossible to ignore.

This creates a systemic blind spot. When we refuse to model agency, we struggle to explain phenomena like the transition from single-celled organisms to multicellular collectives. If we treat cells as passive machines, the alignment of their individual purposes into a single, cohesive entity seems miraculous. If we treat them as agents, the problem becomes a study in how distributed systems align their internal goals.

If we do not start thinking about agency and biology, it just crops up cryptically. We cannot get rid of it. Agency and purpose and goals are intrinsic to biology so let us be clear and honest about that and bring it into the open.

-- Philip Ball

The consequence of this avoidance is a lack of a unified theory. We are currently trying to build agentic AI by simply scaling up models, hoping they will eventually exhibit true agency. Without a formal theory of what agency is, the ability to set and pursue goals in a context-dependent manner, we are guessing at the ingredients rather than engineering the outcome.

Mapping the Future: From Biological Agents to AI

The most significant effect of this discussion is the potential for a naturalized theory of agency. Researchers like Kevin Mitchell and Henry Potter are attempting to strip away the theological and psychological baggage of the term to define it in purely mechanistic, observable terms.

This has implications for AI development. If we can define the requirements for biological agency, the ability to integrate history, internal state, and external context, we can stop treating AI as a black box that might magically become sentient. Instead, we can identify exactly which components are missing. The competitive advantage here lies in moving away from the brute-force approach of bigger models toward an architectural approach that builds in the capacity for genuine, goal-directed improvisation.

Key Action Items

• Audit your machine assumptions: When analyzing complex systems, whether in code, teams, or biology, stop assuming that flexible behavior is just a more complex script. Look for where the system is improvising based on historical context. (Immediate)
• Shift from classification to behavior: Instead of asking if this system is alive or intelligent, ask if this system exhibits agency. This allows you to analyze viruses, AI, and bacteria using the same set of criteria. (Immediate)
• Invest in Agentic architecture: For those in AI and tech, prioritize research into systems that can set and reorient their own goals rather than just optimizing for a static objective function. (12-18 months)
• Adopt the Agency lens for organizational design: If your team is struggling with novel problems, recognize that rigid processes, the hardwired approach, will fail. Build for agency by empowering individuals to integrate context and make decisions, rather than trying to script every outcome. (Over the next quarter)
• Review your foundational literature: Read The Power of Life by Jessica Riskin. Understanding the history of how we have struggled to define life will help you avoid repeating the same reductionist traps in your own field. (Next 3-6 months)