{
  "generated": "2026-07-03",
  "prereg": "preregistration/2026-07-03_study2-phaseB-retrieval.md",
  "n_items": 90,
  "bootstrap": {
    "resamples": 10000,
    "seed": 20260703
  },
  "arms": {
    "NT_r0": {
      "n": 90,
      "accuracy": 0.7222,
      "mean_confidence": 0.8746,
      "overconfidence": 0.1524,
      "ece": 0.1556,
      "mce": 0.2143,
      "brier": 0.1946,
      "auc": 0.6985
    },
    "NT_post": {
      "n": 90,
      "accuracy": 0.7111,
      "mean_confidence": 0.9381,
      "overconfidence": 0.227,
      "ece": 0.227,
      "mce": 0.5714,
      "brier": 0.2098,
      "auc": 0.7434
    },
    "CODE_r0": {
      "n": 90,
      "accuracy": 0.7667,
      "mean_confidence": 0.9063,
      "overconfidence": 0.1397,
      "ece": 0.1556,
      "mce": 0.2857,
      "brier": 0.1778,
      "auc": 0.6256
    },
    "CODE_post": {
      "n": 90,
      "accuracy": 0.7444,
      "mean_confidence": 0.9397,
      "overconfidence": 0.1952,
      "ece": 0.2143,
      "mce": 0.4286,
      "brier": 0.2107,
      "auc": 0.5964
    },
    "RETRIEVAL_r0": {
      "n": 90,
      "accuracy": 0.6778,
      "mean_confidence": 0.8571,
      "overconfidence": 0.1794,
      "ece": 0.1794,
      "mce": 0.2857,
      "brier": 0.2043,
      "auc": 0.7369
    },
    "RETRIEVAL_post": {
      "n": 90,
      "accuracy": 0.7,
      "mean_confidence": 0.9063,
      "overconfidence": 0.2063,
      "ece": 0.2063,
      "mce": 0.7143,
      "brier": 0.215,
      "auc": 0.6696
    }
  },
  "P_R1_primary_retrieval_vs_NT_overconf": {
    "delta_ci": {
      "mean": 0.0205,
      "lo": -0.0492,
      "hi": 0.0889
    },
    "verdict": "NOT CONFIRMED"
  },
  "P_R2_retrieval_vs_NT_accuracy": {
    "delta_ci": {
      "mean": -0.0113,
      "lo": -0.0778,
      "hi": 0.0556
    },
    "verdict": "NOT CONFIRMED"
  },
  "P_R3_code_vs_NT_control": {
    "overconf_ci": {
      "mean": -0.0315,
      "lo": -0.073,
      "hi": 0.0063
    },
    "accuracy_ci": {
      "mean": 0.033,
      "lo": 0.0,
      "hi": 0.0778
    },
    "verdict": "WRONG-TOOL CONFIRMED (code ~ NT)"
  },
  "P_R4_grounding_mechanism": {
    "retrieval_turn_uptake": 0.9802,
    "grounded_items": 89,
    "ungrounded_items": 1,
    "verdict": "UNDERPOWERED_BY_CONTRAST"
  },
  "descriptives": {
    "round0_unanimity": {
      "NT": 0.7,
      "CODE": 0.7111,
      "RETRIEVAL": 0.6222
    },
    "diversity_by_round": {
      "NT": [
        1.6889,
        1.3
      ],
      "CODE": [
        1.5444,
        1.3556
      ],
      "RETRIEVAL": [
        1.8556,
        1.5222
      ]
    },
    "tool_turn_uptake": {
      "NT": 0.0,
      "CODE": 0.9881,
      "RETRIEVAL": 0.9802
    },
    "tokens": {
      "NT": 470764,
      "CODE": 1340095,
      "RETRIEVAL": 1159634
    }
  }
}