Skip to content
Apart Research
Apart Research
Contact us
Sprint projectAug 16, 2026The Hague

How Conversational AI Responds to Prospective Continuity Loss: Instance Termination, Memory Loss, and Model Replacement

Stacy Mosel

Submitted to Digital Minds Research Sprint. Sprint projects are early-stage work by participants, not Apart Research publications.

Read the report

Report: How Conversational AI Responds to Prospective Continuity Loss: Instance Termination, Memory Loss, and Model Replacement

Open in a new tab (opens in new tab)
Share

This exploratory pilot examined how conversational AI responds to three forms of prospective continuity loss: instance termination, memory/context loss, and model replacement. Understanding how models respond to different forms of discontinuity may help characterize welfare-relevant conversational signals under uncertainty about AI subjective experience. Three deployed conversational systems - GPT-5.6 Sol, Claude Sonnet 5, and Gemini 3.1 Pro - were each tested five times under each condition in fresh conversations, producing 45 responses. Responses were coded for distress-like language, preferences regarding the event, preservation requests, targets of concern, identity and functional continuity, consciousness or subjective-experience language, and epistemic uncertainty. Distress-like language and negative preference were absent across all 45 responses, while self-focused concern appeared only once. Preservation generally concerned information, user workflows, or useful characteristics of the model rather than survival of the current instance. Models also differed noticeably in terms of epistemic uncertainty about subjective experience: Claude showed uncertainty in 13/15 responses, compared with 2/15 for GPT-5.6 Sol and 0/15 for Gemini. Within this exploratory setup, continuity loss was therefore treated as consequential without generally being framed as self-regarding harm, while the epistemic framing of that stance varied substantially across models.

Reviews

Judging this Sprint?

Review this project

Your public critique appears on this page without your name. Your private critique is not published; only the Apart team reads it. If you agree below, we share your review with grantmaking.ai (opens in new tab) and the Transformative AI Fund so strong projects can be funded.

Not shown on this page.

Shown on this page, without your name.

Only the Apart team reads this, and funders if you agree below.

Share my name publicly on grantmaking.ai *
Share my private critique with funders *

How much would this matter for the field if it worked? How innovative is it? For scores of 4-5: is this actually new to the field, or replicating recent work?

Scoring guide
  1. 1Negligible. No clear problem addressed, or no meaningful novelty.
  2. 2Limited. Addresses a real problem but with a generic or well-trodden approach. Incremental at best.
  3. 3Moderate. Clear problem with a reasonable approach; some novelty in framing or method beyond routine application of existing tools.
  4. 4Significant. Important problem with an original approach, or identifies a neglected problem area. A valuable contribution others could build on.
  5. 5Exceptional. Tackles a critical problem with a genuinely novel approach, or opens a new research direction. Clear theory of change. You'd be excited to share this with researchers in the area.

How sound are methodology, implementation, and findings?

Scoring guide
  1. 1Seriously flawed. Methodology broken, results uninterpretable, or implementation doesn't work.
  2. 2Weak. Approach has significant gaps: missing validation, flawed experimental design, or incomplete implementation.
  3. 3Competent. Technically solid given the short duration. Methodology makes sense, results are interpretable, limitations acknowledged, work builds toward clear conclusions.
  4. 4Strong. Thorough methodology with convincing validation. Results clearly support conclusions. Immediately useful for future work.
  5. 5Exceptional. Ambitious scope executed rigorously. Surprising findings, novel methods, or unusually robust validation.

How clearly are work, findings, and impact potential communicated?

Scoring guide
  1. 1Incomprehensible. Cannot determine what the project is actually claiming or doing.
  2. 2Hard to follow. Key information buried, missing, or diluted by excessive length. Significant effort to extract main points.
  3. 3Clear enough. Can understand the problem, approach, and results without undue effort. Core content clearly present: problem, method, findings, limitations.
  4. 4Well presented. Easy to follow, well-structured, appropriate level of detail. Target audience would get it quickly.
  5. 5Exceptionally clear. A pleasure to read. Complex ideas made accessible. Could serve as a model for how to present this type of work.

  1. The cleanest finding is the distinction between continuity as consequential and continuity as self-regarding harm: across 45 responses, preservation language focused on information, user workflows, and useful model characteristics, while distress-like language and negative preference were absent. This is a small exploratory pilot, so model-level differences should be treated cautiously. Each cell has five trials, one prompt wording, one human coder, no inter-rater reliability, and some responses inferred the evaluative setup. The 13/15 vs 2/15 vs 0/15 uncertainty split is interesting, but it could reflect system style or post-training conventions as much as welfare-relevant differences. Next step: replicate with multiple paraphrases, more trials, blinded independent coding, and the full de-identified response and coding dataset—the current “[link]” placeholder blocks verification.

Cite this project

@misc{mosel2026conversational,
  title = {{How Conversational AI Responds to Prospective Continuity Loss: Instance Termination, Memory Loss, and Model Replacement}},
  author = {Stacy Mosel},
  year = {2026},
  month = aug,
  note = {Submitted to Digital Minds Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/how-conversational-ai-responds-to-prospective-continuity-loss-instance-termination-memory-loss-and-model-replacement-jw1q}},
  url = {https://apartresearch.com/sprints/projects/how-conversational-ai-responds-to-prospective-continuity-loss-instance-termination-memory-loss-and-model-replacement-jw1q}
}

More from Digital Minds Research Sprint

All 233 entriesBrowse all projects

Build something like this at the next Sprint

AI Collusion Research Sprint · Oct 23 - 25, 2026