Many of the systems with which we are struggling today are very large, multi-agent, socio-technical ecosystems, in which there are purposeful, situated, actor-agents each operating in the environment created by the others.

      To give an adequate account of such an ecosystem, one must make three ‘cuts’ in the space of its being:

1. The ‘cartesian’ cut, a formal model/abstraction of a materially observable world: the basis of the ‘modeling relation’ on which engineering depends, implicit in any actor-agent’s view of itself in its environment.
2. The ‘Heisenberg’ cut, a distinction about that which is observed by the actor-agent based on the particular effects of observation/ measurement, the need for which arises as a consequence of possessing agency, since that agency is both constituted by and constitutive of the environment in which it acts.
3. The ‘endo-exo’ cut, distinguishing the epistemic from the ontic: for example, the behaviour of a system’s users is constrained exogenously by the relationship between the system’s design and its contexts of use.  But insofar as those exogenous constraints are under-determining, the behavior of those users is also endogenously constrained by the way they and other users choose to enact their agency within the ecosystem.
 
      Unfortunately, our current systems modeling and analysis techniques can adequately articulate only the Cartesian cut.

      Projective analysis is an approach to modeling the actor-agents’ models of their relationship to their ecosystem, a particular implementation of which has been developed over the last two decades (referred to as PAN) and is now being taken up by CMU-SEI, providing the means to describe agents dually:

As they model their environment as an organisation of drivers and their ladders of values and effects, and
as they model how they and others work, synchronise their behaviors, and are held accountable.
 
      By eliciting and representing actor-agents’ models of themselves and the environments in which they act, projective analysis gains access to analytical power that extends over the three cuts and yields a stratified matrix from which a graphic account of the actor-agent’s risk exposure may be constructed and presented so that they may evaluate alternative evolutionary strategies – at whatever scale the actor-agent defines itself.