The Universal Optimism of the Self-Evidencing Mind

The Universal Optimism of the Self-Evidencing Mind“:
Karl Friston’s free-energy principle casts agents as self-evidencing through active inference. This implies that decision-making, planning and information-seeking are, in a generic sense, ‘wishful’.
We take an interdisciplinary perspective on this perplexing aspect of the free-energy principle and unpack the epistemological implications of wishful thinking under the free-energy principle. We use this epistemic framing to discuss the emergence of biases for self-evidencing agents. In particular, we argue that this elucidates an optimism bias as a foundational tenet of self-evidencing. We allude to a historical precursor to some of these themes, interestingly found in Machiavelli’s oeuvre, to contextualise the universal optimism of the free-energy principle.

Self-evidencing involves both the generative model and the generative process. This figure is a heuristic for optimistic self-evidencing. The agent has both stronger predictions about good outcomes and updates their beliefs more to good outcomes (green arrows, left). As the agent predicts and updates more to good outcomes, they select actions of engagement (optimistic, green arrow, right) rather than disengagement (pessimistic, red dashed arrow, right) with the world. The actions the agent selects affect the generative process such that their future observations can be dynamically influenced. Although optimistic self-evidencing is constrained by the generative process, the agent’s actions also influence the generative process through the dynamic coupled oscillation process described in the main text. One can imagine the same figure with a pessimistic agent whose predictions and belief updates are stronger for bad outcomes (i.e., a thick red arrow, left) and how this would lead to a different outcome in the dynamic process.
A graphical depiction of an active inference model and the expected free-energy equation formalised in terms of information seeking and reward seeking.

Even if the optimism bias can be modelled in active inference and cast as rational (given a belief about minimising missed opportunity), a more complex picture emerges when viewed in the broadest possible terms of the free-energy principle.
Given the increase in entropy over time the universe is moving toward more disorder. This means that thermodynamic fluctuations are tugging at our (Markov blanket) boundaries at all times. Concretely, this means that, every time an agent has arrived at a belief about something on the basis of their self-evidencing, they will learn that this is a kind of pyrrhic victory, since dissipative forces will soon undermine that belief and turn the positive outcome into a negative one that is less consistent with the agent’s belief.
Put differently, as the current posterior becomes the prior for the next inference, agents should come to appreciate that all priors decline over time, mandating new belief formation (previously, we have leveraged this idea in conceptual and active inference models of binocular rivalry). This persistent occurrence of negative outcomes suggests, at least initially, that, over appropriate time horizons, we really should be pessimists. Our opportunities for self-evidencing are, after all, limited by all the uncertainty we encounter in the shape of dissipative forces that, in the end, will literally tear us apart (that is, irreparably damage our organismic boundaries or Markov blankets).
No one lives forever, and societies and species vanish, having existed only as tiny blips on the radar of the universe. Puzzlingly, in spite of this prima facie case for pessimism, we, and many animals, are staunchly optimistic.

It may be, therefore, that the optimism bias observed in the everyday behaviour of many species is an expression of the fundamental tenet of the free-energy principle, namely that our existence is inextricably bound up with a belief that the world benevolently affords our existence—that the world offers us opportunities to seize in spite of its consistently observed inhospitality.
This might be regarded as instrumentally rational because it keeps us going in the medium term, but on a more cosmic time scale, it does begin to look like wishful thinking in the pejorative sense.

Leave a comment