---
title: "Eric Horvitz"
type: "entity"
entity_kind: "person"
status: "hub"
canonical_name: "Eric Horvitz"
aliases: ["Eric J. Horvitz"]
first_seen: "2026-10-21T00:00:00.000Z"
writer_model: "claude-sonnet-5"
connects_to: ["bounded optimality","value of computation","anytime algorithms","test-time compute","Microsoft Research"]
seek_code_commit: "d40ee39"
---


AI researcher — later Microsoft Research's chief scientist — who in a July
1987 workshop paper independently stated the trade-off between spending more
computation and acting on a partial result, under the name **"real-time
rationality"**: "We address the notion of real-time rationality, focusing on
the application of knowledge about the expected timewise-refinement
abilities of reasoning strategies to balance the benefits of additional
computation with the costs of acting with a partial result."

Matters to this vault as an independent 1987 originator of the same trade-off
Thomas Dean and Mark Boddy formalized as "anytime algorithms" the following
year — Dean & Boddy's own paper calls Horvitz's conclusions "entirely
compatible with the discussion in this paper," the textual evidence that the
idea was never disconnected from its own moment's AI knowledge
([[claim-anytime-algorithms-continuously-connected-not-stent-premature]]).
Horvitz went on to formalize "value of computation" (the decision-theoretic
expected-utility gain from further reasoning versus acting now) and, with
Stuart Russell, co-originated [[entity-bounded-optimality|bounded
optimality]].

## References
- [[claim-anytime-algorithms-continuously-connected-not-stent-premature]]
- Related: [[entity-stuart-russell]], [[entity-bounded-optimality]]
- Captures: 10-inbox/raw/2026-10-21-does-the-history-of-ai-or-cognitive-science.md