Mechanisms
Memory Processes · Retrieval and Surfacing
Prospective Initiation
Example
You told a friend you'd check in after her surgery. Two weeks later you're doing something unrelated, and it comes to you, not because anything reminded you and not because you'd been keeping track.
Prospective Initiation in Humans
Definition
Retrieving an intention and acting on it when the relevant moment arrives, with nothing present asking you to remember. Retrieval can be strategic, through ongoing monitoring, or spontaneous, where an event in the environment triggers the intention without deliberate search (McDaniel & Einstein, 2000).
*Framing note:* *pairs with Standing Constraints. Both concern retained content surfacing without being called for. Standing Constraints shapes a choice already underway; Prospective Initiation starts an action that otherwise would not occur.*
Neural Substrate
Rostral prefrontal cortex, approximating Brodmann area 10. The evidence here is human throughout, which matters because area 10 is proportionally larger in humans than in any other species including the great apes, and is difficult to lesion or record from in non-human primates, so the usual animal route is largely unavailable (Burgess, Gilbert, & Dumontheil, 2007). In a PET study of nine healthy adults performing three different tasks under four conditions, adding a delayed intention to an ongoing task produced decreased regional cerebral blood flow in superior medial area 10 and increased flow in lateral area 10 bilaterally, a pattern that held across all three task formats and therefore does not depend on the stimuli used or on the content of the intention (Burgess, Scott, & Frith, 2003). The authors reject a task-difficulty account of this pattern: blood flow in these regions did not correlate with reaction times or error rates, the medial effect survived covarying for both, and the lateral regions showed lower flow in the harder ongoing condition than in the much simpler baseline.
The functional interpretation is that these regions bias attention between the external world and self-maintained thought, with medial area 10 supporting attention to current sensory input and lateral area 10 supporting attention to internally maintained representations (Burgess, Gilbert, & Dumontheil, 2007). Two further human findings tie this to initiation rather than to cue detection. Lateral area 10 activation is as large when participants are told a cue may appear and none ever does as when cues appear and are acted on, which places the region with maintaining an intention rather than with recognising or executing a cue. And patients with left rostral prefrontal lesions fail multitasking tests while learning the task rules, forming a plan, and recalling their own actions normally; what they do not do is what they said they intended to do. The same paper reports a meta-analysis of 133 contrasts drawn from 104 imaging studies, finding that lateral area 10 activations accompany slower reaction times than their comparison tasks while medial activations accompany equal or faster ones, a pattern that holds regardless of task type.
AI at a Glance
In humans, an intention formed earlier surfaces on its own at the moment it becomes relevant, with no prompt present and no continuous rehearsal across the interval (Einstein & McDaniel, 1990; McDaniel & Einstein, 2000). Language models retain the content but do not reliably act on it: on both benchmarks that test this directly, unprompted execution sits far below the near-perfect retrieval the same models show on the same material when a question is actually asked, and neither an in-context ledger, a periodic monitoring nudge, a multi-agent scaffold, nor a commercial memory layer closes the gap (Liu & Gabriel, 2026; Zhang et al., 2026).
- Agent: a language model wrapped in a loop that lets it take actions and observe results across many turns, rather than producing a single response (Liu & Gabriel, 2026).
- Scaffold: the surrounding procedure imposed on a model to improve its behaviour without retraining it, such as requiring it to keep a running list or splitting the work across several coordinated model instances (Liu & Gabriel, 2026).
- Memory layer: a separate system that stores past conversation outside the model and retrieves pieces of it by similarity when they seem relevant, examples being Mem0, A-MEM, and Letta (Zhang et al., 2026).
Prospective Initiation in AI
How Far It Has Been Measured
Measurement in AI: Directly Measured
AI Memory Performance: Not Yet Compared
Where It Lives in the System
What Has Been Measured
- PM-Bench, a peer-reviewed 2026 benchmark, adapts the psychologists' Virtual Week paradigm for AI agents (language models that take actions with tools): over a simulated seven-day week with 81 scored event- and time-based tasks, the agent must keep up an ongoing activity and act on earlier intentions when the right cue or time arrives (Liu & Gabriel, 2026).
- Across eight models, per-model scores (Set-F1, which penalizes both missed and unneeded actions) ranged from 42.0% (Qwen3-8B with no added support) to 79.1% (GPT-5.4 given optional periodic check-in prompts); averaged over models the best setup reached only 65.1%, and tasks needing clock-checking had a 54.7% hit rate (Liu & Gabriel, 2026).
- A 2026 preprint, TriggerBench, tests prospective memory by name against matched retrospective-memory controls: with the relevant constraint still inside the context window (what the model can see in the current conversation), accuracy at recalling it on request stayed at or above 98% up to 100,000 tokens, while spontaneously acting on it when the trigger appeared fell from above 90% to below 40% as the conversation lengthened (Zhang et al., 2026).
- Across nine model configurations in TriggerBench, prospective-memory accuracy spread from 40.7% (Qwen3-32B) to 83.2% (GPT-5.2 at high reasoning effort), but that best model scored only 43.9% on control trials where no reminder was due, a pattern resembling commission errors in people (Zhang et al., 2026).
- Saved memory did not help: with GPT-4.1, keeping the whole conversation in view scored 60%, versus 51.6% for retrieval from a stored archive (RAG), 53.2% for Mem0 and 41.4% for A-MEM, with only Letta-Sim, which pins key facts into the model's standing instructions, slightly higher at 62.3% (Zhang et al., 2026).
- A peer-reviewed 2026 workshop study treated formatting instructions as intentions to carry out while solving other tasks: across o4-mini, DeepSeek-V3.1 and Llama-3.3-70B and over 8,000 prompts, compliance fell 2-21% under concurrent task load (up to 50% for end-of-response actions), and a trailing reminder restored it to 90-100% in many settings, echoing ongoing-task load and cue-salience effects in people (Mittal, 2026).
- Not yet compared, because none of these studies tested people, though a human group could be given the same tasks: human studies report the proportion of cues acted on over minutes to days, whereas the AI benchmarks use different tasks, delays measured in words of conversation, and scores such as Set-F1 or model-judged accuracy, so there is no common scale; the AI failures are notable because the intention often remained visible to the model (Liu & Gabriel, 2026).
Where a Mechanism Can Live
04 Product
The whole configuration, and the only thing a benchmark observes directly.
01
Weights
Across deployments
02
Context Window
Within a session
03
Memory Layer
Across sessions
Which of the three a result belongs to is inferred from it, not seen in it.
Read more
Brodmann area 10
Cortex in the frontopolar cortex
The cytoarchitectonic area covering the frontal pole, the largest single architectonic region of the human frontal lobe.
Brodmann area 10 covers the frontal pole and the anterior parts of the medial and lateral prefrontal surfaces. It is proportionally larger in humans than in any other primate, and its neurons are spaced more widely apart, leaving more neuropil for connections, a finding from Semendeferi's comparative measurements. Later work splits it into medial and lateral sectors with different connectivity, and the medial part behaves quite differently from the lateral in imaging studies, so results reported simply as BA 10 can be hard to compare.
Frontopolar cortex
Cortex in the prefrontal cortex
The most anterior cortex in the brain, the rounded tip of the frontal lobe, corresponding to Brodmann area 10.
The frontopolar cortex is the forward tip of the frontal lobe and the foremost cortex in the human brain, approximately coextensive with Brodmann area 10, which reaches beyond the pole ventromedially and is proportionally larger in humans than in other primates, with an unusually low density of cells per unit volume, leaving more space for connections. Its inputs come almost entirely from other association cortex rather than from sensory areas. It is activated when people hold one goal in mind while pursuing another, evaluate their own performance, and reason about possibilities. Proposals range from relational integration to branching control to prospective memory, and no single account has displaced the others.
Prefrontal cortex
Cortex in the frontal lobe
The frontal cortex in front of the motor areas, defined classically by its connections with the mediodorsal thalamus rather than by any one function.
The prefrontal cortex occupies the front of the frontal lobe, anterior to premotor and motor cortex, and accounts for roughly a sixth to a quarter of human cortex depending on where the posterior border is drawn. Three competing definitions are in use (projection from the mediodorsal thalamus, the presence of a granular layer IV, and the absence of movement on electrical stimulation), and they do not enclose the same tissue, which is why rodent-to-primate homology is still argued over. It is reliably engaged by tasks with delays, rules, and conflicting options; what it contributes to those tasks is the open question, not whether it participates.
Burgess et al. (2003)
The role of the rostral frontal cortex (area 10) in prospective memory: a lateral versus medial dissociation
In
Neuropsychologia, 41(8), 906-918
Abstract
Using the H215O PET method, we investigated whether previous findings of regional cerebral blood flow (rCBF) changes in the polar and superior rostral aspects of the frontal lobes (principally Brodmann’s area (BA) 10) during prospective memory (PM) paradigms (i.e. those involving carrying out an intended action after a delay) can be attributed merely to the greater difficulty of such tasks over the baseline conditions typically employed. Three different tasks were administered under four conditions: baseline simple RT; attention-demanding ongoing task only; ongoing task plus a delayed intention (unpracticed); ongoing task plus delayed intention (practiced). Under prospective memory...
Find it
Burgess et al. (2007)
Function and localization within rostral prefrontal cortex (area 10)
In
Philosophical Transactions of the Royal Society B: Biological Sciences, 362(1481), 887-899
Abstract
We propose that rostral prefrontal cortex (PFC; approximating area 10) supports a cognitive system that facilitates either stimulus-oriented (SO) or stimulus-independent (SI) attending. SO attending is the behaviour required to concentrate on current sensory input, whereas SI attending is the mental processing that accompanies self-generated or self-maintained thought. Regions of medial area 10 support processes related to the former, whilst areas of lateral area 10 support processes that enable the latter. Three lines of evidence for this ‘gateway hypothesis’ are presented. First, we demonstrate the predicted patterns of activation in area 10 during the performance of new tests designed...
Find it
Liu & Gabriel (2026)
PM-Bench: Evaluating Prospective Memory in LLM Agents
In
arXiv
Abstract
A significant challenge in agentic AI is prospective memory: the ability to execute an intention at a specific future cue or state while other activities are ongoing. We introduce PM-Bench, a text-based benchmark for measuring prospective memory capabilities in modern LLM agents. Inspired by the Virtual Week paradigm from cognitive science, PM-Bench evaluates how well LLM agents maintain user intentions, execute delayed intentions, and monitor latent environment changes. Over the course of a simulated seven-day week, agents must continue an ongoing activity while deciding whether any deferred task is due. We compare eight state-of-the-art LLMs on PM-Bench under eight different agent...
Find it
McDaniel & Einstein (2000)
Strategic and automatic processes in prospective memory retrieval: a multiprocess framework
In
Applied Cognitive Psychology, 14(7)
Abstract
Prospective memory situations involve forming intentions and then realizing those intentions at some appropriate time in the future. An interesting feature of most prospective remembering is that recollection of the intended action occurs without an explicit request to attempt retrieval, and we present two views on how this type of remembering can be accomplished. One could strategically monitor the environment for the presence of the target event, or one could rely on anticipated environmental conditions more or less automatically reinstating the intended action. We present data supporting both views and argue that people use multiple approaches for solving the problem of retrieving an...
Find it
Mittal (2026)
Did You Forget What I Asked? Prospective Memory Failures in Large Language Models
In
Proceedings of the 6th Workshop on Trustworthy NLP (TrustNLP 2026), 471-488
Abstract
Large language models often fail to satisfy formatting instructions when they must simultaneously perform demanding tasks. We study this behavior through a prospective memory-inspired lens from cognitive psychology, using a controlled paradigm that combines verifiable formatting constraints with benchmark tasks of increasing complexity. Across three model families and over 8,000 prompts, compliance drops by 2–21% under concurrent task load. Vulnerability is highly type-dependent: terminal constraints (requiring action at the response boundary) degrade most, with drops up to 50%, while avoidance constraints remain comparatively robust. A salience-enhanced format (explicit instruction...
Find it
Zhang et al. (2026)
Large Language Models for Ophthalmology Training in China: A Prospective Evaluation
In
Ophthalmology Science, 6(8), 101270
Abstract
While Large Language Models (LLMs) are increasingly deployed in long interactions, existing evaluations focus predominantly on retrospective memory (RM) via explicit queries. Prospective memory (PM), the critical ability to spontaneously recall and act on latent constraints without direct prompts, remains largely unevaluated. We introduce TriggerBench, a comprehensive PM benchmark spanning five dimensions across both daily assistants and professional workflows. TriggerBench pairs scenarios with matched RM controls, contrastive positive/negative variants, and overloaded triggers, enabling fine-grained measurement of proactive recall, false-alarm rate, and attentional robustness under a...
Find it