We use genetic programming evolved networks, vector fields, and signal processing to study time varying-exposures where trust is implied (e.g. a conversion event from attention flow to financial commitment). The datasets are behavioral finance time series (from on-chain data, such as fees, and off-chain data, such as clickstreams), which we use to elaborate on various complexity metrics of causality, through the creation parametric network graphs. We discuss the related methods and applications and conclude with the notion of social memory irreversibility and value by memory as useful constructs that take advantage of the natural fact of the existence of trust asymmetries, that can be operationalized by embedded AIs that use distributed ledgers both as the substrate of their intelligence and as social computers.   By being context-aware, those intelligent agents are able to intervene in problematic stressors and contribute to minimizing network fragility.
Keywords: systemic risk, behavioral finance, economic complexity, evolutionary computation, computational trust, the blockchain,  cryptocurrencies, market microstructure, reality mining.