Indice
A new category for understanding the relationship between human and artificial in the algorithmic era
We live in an age where artificial intelligence is no longer just a tool in our hands, but an active mediator that co-constitutes our actions, our choices, even the way we perceive and inhabit the world. Yet we continue to describe this reality with inadequate categories: on one hand, the reduction of AI to a mere passive instrument; on the other, its hyperbolic elevation to an autonomous entity threatening to replace the human.
Participated Agency emerges as a third way, an intermediate category that overcomes this false dichotomy and offers more precise conceptual tools for understanding the ontological specificity of contemporary algorithmic systems.
Beyond the Instrumental Paradigm
The traditional conception of technology as a neutral tool, derived from the Aristotelian-Thomistic doctrine of instrumental causality, presupposes that artifacts act exclusively by virtue of the principal agent, passively transmitting human intention without modifying it. However, empirical observation of contemporary socio-technical systems reveals a far more complex reality.
A machine learning algorithm doesn’t simply execute the programmer’s commands: it learns from data that no human being has ever analyzed in their totality, establishes correlations that no one had foreseen, produces decisions that surprise its own creators. Technological artifacts don’t merely transmit human intention: they mediate it, transform it, redefine it through their own operational logics.
At the same time, the perspective of radical symmetry proposed by Actor-Network Theory (ANT) – which dissolves every ontological distinction between human and non-human actants – risks erasing precisely the priority of human agency that constitutes the foundation of moral responsibility.
The Three Levels of Participated Agency
Participated Agency resolves this tension by articulating itself across three levels of increasing mediative complexity:
1. Instrumental Agency: The Constitutive Mediator
At the most elementary level, we find systems that creatively participate in action through relational processing capacities, but maintain substantially classical causality. A calculator performing arithmetic operations, a search engine with predefined keywords, an industrial control system with fixed parameters: here the artifact transmits the efficacy of the human agent without significantly altering its nature.
2. Adaptive Mediative Agency: Partial Transformation
At an intermediate level, we find systems that introduce mediative elements that partially transform the original intention. A recommendation algorithm doesn’t merely process what the user explicitly requests: it interprets their preferences, contextualizes them relative to the behaviors of similar users, proposes unconsidered content. Human intention is mediated through algorithmic logics that respecify and enrich it, introducing their own criteria that orient the final action in potentially divergent directions from the original intention, while maintaining ontological subordination to human intentionality.
3. Emergent Co-constitutive Agency: Generative Participation
At the most complex level, the artificial system becomes co-constitutive of human action itself. A high-frequency algorithmic trading system doesn’t simply execute pre-constituted human strategies: it generates emergent strategies that redefine the very conditions of the market. Its effects retroact on the environment in which it operates, creating new conditions that require further adaptations in both algorithms and human operators. The resulting action is the product of genuine co-constitution between human intentionality and algorithmic logics.
However – and this is the crucial point – such co-constitution remains asymmetric: algorithms always operate within spaces of possibility defined by fundamental human choices that determine architectures, objectives, ethical constraints, and modes of control.
A Concrete Example: Google Maps
Consider our daily interaction with Google Maps. When we request a route to a destination, the algorithm processes the request considering traffic, distances, user preferences and returns an optimal path. During the journey, we notice intense traffic not yet reported and autonomously decide to change route, simultaneously reporting the incident. Google Maps receives this information, updates its data and automatically recalculates alternative routes for other users.
This process perfectly illustrates participated agency:
Ontological dependence: without the human request for a route, the algorithm would have no operational reason. The ultimate purpose remains entirely human.
Active mediation: the algorithm processes variables that the user couldn’t simultaneously consider (global traffic conditions, historical times, road closures, optimization algorithms).
Emergent results: it suggests routes that no programmer had specifically coded for that particular combination of circumstances. It can surprise the user by proposing roads never considered.
Ultimate subordination: it always operates in view of purposes established by human intelligence, processes data generated by human decisions, and its causal efficacy is always exercised as mediation of pre-existing human projects.
The example dissolves the supposed contradiction between transformative mediation and ontological subordination.
Implications for Moral Responsibility
Recognition of participated agency has decisive consequences for the distribution of moral responsibility. If artificial systems are genuine mediators that transform human action, responsibility can no longer be conceived according to the traditional linear model.
Responsibility must be articulated as distributed responsibility in networks of agency, considering:
- Design responsibility (who designs the architecture)
- Implementation responsibility (who develops the system)
- Deployment responsibility (who implements it in specific contexts)
- Supervision responsibility (who monitors functioning)
- Governance responsibility (who defines policies)
This distribution doesn’t equate to dilution: each level maintains its specificity while operating in constitutive relation with the others.
From Analysis to Praxis
Participated Agency is not academic speculation, but a necessary foundation for responsibly orienting the development and use of artificial intelligence. Recognizing that AI configures an existential environment – and not simply a tool to be regulated – obliges us to radically rethink our ethical, professional, and educational strategies.
This category allows us to:
- Overcome naively enthusiastic or a priori condemning approaches toward AI
- Recognize digital structures of sin (technological configurations that systematically orient toward injustice)
- Discern possible digital structures of grace (algorithmic mediations that favor human flourishing)
- Elaborate more sophisticated evaluation criteria for algorithmic systems
- Design technologies that respect human dignity and autonomy
Toward a Conscious Digital Culture
Participated Agency invites us to a new literacy: not only knowing how to use technologies, but understanding how they use us, how they mediate our perception of reality, how they co-constitute our decision-making processes, how they structure our spaces of possibility.
In an age where algorithms have become environment – invisible yet omnipresent, silent yet performative – we need conceptual categories that allow us to inhabit this space critically and creatively. Participated Agency represents a contribution in this direction: an attempt to think technology with the rigor that the complexity of our contemporary condition requires.