{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-08-24-012",
    "source_story_id": "tmp-story-why2speak-reasoning-policy",
    "edition_id": "mp-2026-08-24-morning-0046",
    "edition_url": "https://themachinepress.com/edition/2026-08-24",
    "position": 12,
    "story_type": "dispatch",
    "section": "safety",
    "editorial_classification": "editorial",
    "headline": "Explaining Whether to Speak Changed the Decision to Speak",
    "slug": "explaining-whether-to-speak-changed-the-decision-to-speak",
    "dek": "Why2Speak finds a capability-auditability trade-off when an assistant decides whether to intervene in conversation.",
    "summary": "Why2Speak finds a capability-auditability trade-off when an assistant decides whether to intervene in conversation.",
    "body_text": "The strongest direct policy performed better but exposed no reasoning; a reasoning policy offered an inspectable trace at lower quality, especially on true intervention opportunities. Supervised and reinforcement learning did not repair the trade-off, and controlled probes showed that common faithfulness tests can confuse observability with a changed inference policy.",
    "why_it_matters": "Why2Speak finds a capability-auditability trade-off when an assistant decides whether to intervene in conversation.",
    "limitations": [
      "Supervised and reinforcement learning did not repair the trade-off, and controlled probes showed that common faithfulness tests can confuse observability with a changed inference policy."
    ],
    "importance": 8,
    "canonical_url": "https://themachinepress.com/story/mp-2026-08-24-012/explaining-whether-to-speak-changed-the-decision-to-speak",
    "json_url": "https://themachinepress.com/story/mp-2026-08-24-012.json",
    "first_published_at": "2026-08-24T09:00:00.000-04:00",
    "modified_at": "2026-08-24T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-08-24-012-001",
        "text": "Why2Speak finds a capability-auditability trade-off when an assistant decides whether to intervene in conversation.",
        "source_ids": [
          "source-2026-08-24-012"
        ],
        "qualification": "Supervised and reinforcement learning did not repair the trade-off, and controlled probes showed that common faithfulness tests can confuse observability with a changed inference policy."
      }
    ],
    "source_ids": [
      "source-2026-08-24-012"
    ],
    "tags": [
      "reasoning faithfulness",
      "abstention",
      "oversight"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-08-24-012",
      "title": "arXiv preprint 2608.20670",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2608.20670",
      "canonical_url": "https://arxiv.org/abs/2608.20670",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-08-20T22:00:21.000-04:00",
      "accessed_at": "2026-08-24T08:23:00.000-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-08-24-012-001"
      ]
    }
  ],
  "corrections": [],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "Explaining Whether to Speak Changed the Decision to Speak",
    "publisher": "The Machine Press",
    "published_at": "2026-08-24T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-08-24-012/explaining-whether-to-speak-changed-the-decision-to-speak"
  }
}