digg/ai
5h ago

Agent Actions Function as Causal Interventions in Posterior Inference

——0——
Original post
NandoDF
Nando de Freitas#29@NANDODF

Causal vs observational agency. Agent actions (a) should be treated as interventions, not as evidence for hypotheses (p). Actions by other agents or tool outputs are evidence (o).

5:32 AM · May 17, 2026 View on X
ORIGINAL POSTNando de Freitas#29Nando de Freitas@NANDODF

Causal vs observational agency. Agent actions (a) should be treated as interventions, not as evidence for hypotheses (p). Actions by other agents or tool outputs are evidence (o).

12:32 PM · May 17, 2026 · 4.4K Views
REPLYNando de Freitas#29Nando de Freitas@NANDODF

In the language of LLM losses:

Nando de FreitasNando de Freitas@NandoDF

Causal vs observational agency. Agent actions (a) should be treated as interventions, not as evidence for hypotheses (p). Actions by other agents or tool outputs are evidence (o).

12:32 PM · May 17, 2026 · 4.4K Views
12:50 PM · May 17, 2026 · 1.4K Views
Agent Actions Function as Causal Interventions in Posterior Inference · Digg