Rebuild: add trajectories (passive activations, decision record), impressions.slot, actions.step; fix step column (see BUILD_LOG.md erratum)

BUILD_LOG.md

@@ -0,0 +1,96 @@
+# Build log — 2026-09-09
+
+Rebuild of all tables from the 1,296 source databases, adding `trajectories.parquet`,
+`impressions.slot`, `actions.step`, and correcting the `step` column.
+
+## Commands
+
+```
+python build_divdata.py       --out divdata/                         # 6 core tables, ~2h15 single process
+python build_trajectories.py  --out divdata/ --workers 16 --patch-steps   # trajectories + step back-fill
+```
+
+Source: `<run>.db` (SQLite, OASIS schema) + `<run>.metadata.json` (agent → model map)
++ `<run>_chats.jsonl` (one row per activation: step, status, attempts with the model's
+response and tool call) + `<run>_reasoning.jsonl` (thinking channel, where exposed).
+1,384 databases found; 88 skipped (no metadata sidecar, or an in-progress checkpoint).
+
+## Output
+
+| table | rows | size | step non-null |
+|---|---:|---:|---:|
+| runs | 1,296 | 34 KB | — |
+| agents | 28,938 | 24 KB | — |
+| posts | 89,171 | 23.8 MB | 90.2% |
+| comments | 652,588 | 91.8 MB | 91.8% |
+| impressions | 14,020,992 | 1.2 MB | 93.3% |
+| actions | 741,219 | 6.1 MB | 91.7% |
+| trajectories | 748,285 | 169.6 MB | 89.5% (of feed-fetch activations) |
+
+Row counts of the six core tables are identical to the previous release; only `step`
+values, and the new columns, changed.
+
+## Trajectories build log (`_trajectories_build_log.json`)
+
+```
+runs 1296        no_chats 41 (no sidecar → timestamps and actions only, no step/decision)
+agents 28847     agents_aligned 26786 (92.9%)   agents_multi 339 (see below)
+activations 748285 = 741219 actions + 7066 passive (fetched the feed, no action landed)
+status: ok 670505 · exhausted 2263 · exception 79 · null 75438 (unaligned agents)
+with_tool_call 642083 · with_response_text 258949 · with_reasoning 284788
+windows_multi 3247 (extra_in_window rows)
+```
+
+Validation:
+- every action in `actions.parquet` is present in `trajectories` (741,219 = 741,219);
+- every feed-fetch window holds at most one action in all but 16 runs;
+- where `trajectories.step` and the action-count alignment of `posts`/`comments` both give
+  a value, they agree on 624,776 / 624,820 rows (99.993%); all 44 disagreements are in
+  5 of those 16 runs;
+- `--patch-steps` back-filled step from the feed-fetch alignment where the action-count
+  alignment had NULL: posts 15,060 → 8,701 NULL; comments 91,744 → 53,403; actions
+  106,264 → 61,564. Agreement on rows both methods covered: 99.995 / 99.993 / 99.993%.
+
+### The 16 runs with two actions in one feed-fetch window
+
+In decreasing number of extra actions: `archive_20260516_1003/oasis_v9_hetero_late` (2,468 —
+an early v9 run whose harness allowed multiple actions per activation), `sweep_olmo_mag_s42`
+(98), `sweep_mag_olmo_s42` (97), `tetratic/tetra_s271_r4` (82), `tetratic/tetra_s271_r3` (81),
+`attractor_rl/eval_runs_triadic/baseline/eval_baseline_s43` (72), `…_s42` (66),
+`tetratic/tetra_s42_r5` (57), `sweep_glm47_glm47_qwe_s137` (51), `sweep_glm47_glm47_qwe_s42`
+(36), `sweep_tri_glm_qwe_gem_s999` (36), `sweep_tri_glm_qwe_gem_s271` (32),
+`tetratic/tetra_s42_r6` (31), `sweep_tri_gem_mag_glm_s999` (20), `sweep_tri_gem_mag_glm_s271`
+(16), `sweep_tri_gem_qwe_glm_s42` (4). For an agent with any such window the refresh alignment is not
+trustworthy, so its `step`, `status` and decision fields are NULL in `trajectories`
+(339 agents); its actions are all present, extras flagged `extra_in_window=True`, and
+`refreshed_at` / `acted_at` are exact. Step coverage in these runs is 0–58%. Prefer
+`created_at` ordering there.
+
+## Erratum in the previous release (published 2026-09-04)
+
+`build_divdata.py` assigned step by iterating all of an agent's posts, then all of its
+comments, while the alignment index assumed one time-ordered stream. Consequence, measured
+by rebuilding and comparing on the same rows:
+
+| table | rows compared | old step == new step |
+|---|---:|---:|
+| posts | 74,111 | **38.9%** |
+| comments | 560,844 | **49.2%** |
+| impressions | per viewer | 0% of viewers had the same step set — every value shifted by one activation (the step-0 intro post has no feed fetch, and was counted as one) |
+
+Agents that only ever commented, and the step-0 intro posts, were correct; everything else
+was scrambled. No other column was affected. `actions.step` was NULL throughout in the old
+release. Nothing in the accompanying paper used `step`.
+
+## Decision-record coverage by model (feed-fetch activations)
+
+| model | activations | response_text | reasoning | tool_call | passive |
+|---|---:|---:|---:|---:|---:|
+| gpt-oss-20b | 247,489 | 0.4% | 4.8% | 94.4% | 0.3% |
+| Magistral-Small-2509 | 149,666 | 86.2% | 91.7% | 91.3% | 1.1% |
+| Qwen3-32B | 102,546 | 44.6% | 44.7% | 90.2% | 0.3% |
+| GLM-4-32B-0414 | 54,430 | 0.0% | 0.0% | 95.7% | 0.6% |
+| gemma-4-31B-it | 39,744 | 93.0% | 93.4% | 92.7% | 1.2% |
+
+`tool_call` < 100% because unaligned agents carry no decision record. gpt-oss and GLM-4
+never emit visible deliberation.

README.md

@@ -27,6 +27,8 @@ configs:
     path: impressions.parquet
   - split: actions
     path: actions.parquet
+  - split: trajectories
+    path: trajectories.parquet
 ---
 
 # divdata
@@ -34,7 +36,7 @@ configs:
 Every simulation run behind our heterogeneous-LLM social-simulation work, consolidated into
 one dataset indexed by `run_id` and `step`.
 
-**1,296 simulation runs · 89,171 posts · 652,588 comments · 14,020,992 impressions · 32 model variants.**
+**1,296 simulation runs · 89,171 posts · 652,588 comments · 14,020,992 impressions · 748,285 agent activations · 32 model variants.**
 
 Agents with distinct personas post and comment on a shared message board built on
 [OASIS](https://github.com/camel-ai/oasis). Each agent is driven by one of ~10 different
@@ -64,7 +66,7 @@ cpp = (coms[coms.is_cross_model].groupby("model").size()
 
 ## Tables
 
-All six files share `run_id` as the primary key. `step` is the simulation tick.
+All seven files share `run_id` as the primary key. `step` is the simulation tick.
 
 | file | rows are | key columns |
 |---|---|---|
@@ -72,8 +74,9 @@ All six files share `run_id` as the primary key. `step` is the simulation tick.
 | `agents.parquet` | one agent in one run | `run_id`, `agent_id`, `model`, `persona`, `join_step` |
 | `posts.parquet` | one authored post | `run_id`, `post_id`, `agent_id`, `model`, `step`, `content` |
 | `comments.parquet` | one authored comment | `run_id`, `comment_id`, `post_id`, `agent_id`, `model`, `step`, `content`, `is_cross_model` |
-| `impressions.parquet` | one post shown to one agent | `run_id`, `step`, `viewer_agent_id`, `post_id` |
-| `actions.parquet` | one logged action | `run_id`, `agent_id`, `action`, `created_at` |
+| `impressions.parquet` | one post shown to one agent | `run_id`, `step`, `viewer_agent_id`, `post_id`, `slot` |
+| `actions.parquet` | one logged action | `run_id`, `agent_id`, `step`, `action`, `created_at` |
+| `trajectories.parquet` | one agent activation, active **or passive** | `run_id`, `agent_id`, `activation_idx`, `step`, `refreshed_at`, `action`, `post_id`, `tool_args`, `response_text`, `reasoning` |
 
 `model` is denormalised onto `posts` and `comments` so the common analyses need no join.
 `is_cross_model` is true when the commenter's model differs from the post author's model —
@@ -81,21 +84,31 @@ this is the engagement signal the work is about.
 
 ## Two things to understand before using this
 
-**1. `step` is reconstructed, and ~15% of rows do not have one.**
+**1. `step` is reconstructed, and ~9% of rows do not have one.**
 
 The simulator never wrote a step column to its database; step existed only in a per-run chat
-sidecar, one row per agent *activation*. We recover it by aligning, per agent, the k-th
-successful activation with the k-th action in time order. Where the two counts agree the step
-is **exact**. Where they disagree — typically one agent in a run off by a single action — we
-set `step = NULL` rather than guess. Check `runs.step_coverage` (0–1) for the fraction of a
-run's actions carrying an exact step. Median across runs is **0.97**; overall **83.1% of
-posts and 85.9% of comments** carry an exact step. `runs.step_alignment_ok` is true only
-when *every* agent in a run aligned perfectly, which holds for 632 of 1,296 runs — a single
-off-by-one agent flips it false, so prefer `step_coverage` over this flag.
-
-If you need strict step semantics, filter to `step.notna()`, or to runs with
-`step_coverage == 1.0`. For time-ordering within a run, `created_at` is always present and
-always monotone.
+sidecar, one row per agent *activation*. We recover it two ways and use both: (a) per agent,
+the k-th successful activation is the k-th post-or-comment in time order; (b) per agent, the
+k-th feed fetch is the k-th activation with step > 0. Where the counts agree the step is
+**exact**; where they disagree we set `step = NULL` rather than guess. Where both methods
+give a value they agree on 99.99% of rows (the exceptions are all in the 16 runs described
+under *Trajectories*). Coverage: **90.2% of posts, 91.8% of comments, 91.7% of actions,
+93.3% of impressions** carry an exact step. `runs.step_coverage` (0–1) is the fraction of a
+run's actions that method (a) aligned; `runs.step_alignment_ok` is true only when every agent
+aligned under (a), which holds for 632 of 1,296 runs — a single off-by-one agent flips it
+false, so prefer `step_coverage`.
+
+If you need strict step semantics, filter to `step.notna()`. For time-ordering within a
+run, `created_at` is always present and always monotone.
+
+> **Erratum (2026-09-09).** Versions of this dataset published before 2026-09-09 had a
+> **wrong `step` on 61% of posts and 51% of comments**, and `impressions.step` was off by
+> one activation. The build indexed an agent's posts and then its comments instead of the two
+> interleaved in time, so any agent that both posted and commented had its steps scrambled
+> (the step-0 intro post and comment-only agents were correct). `created_at` and every
+> non-step column were unaffected. If you used `step` from an earlier download, re-download.
+> The fix is in `build_divdata.py` in the code release, and `BUILD_LOG.md` here has the
+> before/after numbers.
 
 **2. Impressions are exposure, and exposure is extremely unequal.**
 
@@ -108,9 +121,69 @@ engagement per post.
 
 In the source databases each feed refresh embedded the full text of all 20 slate posts,
 which was ~48% of every file. Here impressions are normalised to
-`(run_id, step, viewer_agent_id, post_id)` and post text is stored once — about a 150×
+`(run_id, step, viewer_agent_id, post_id, slot)` — `slot` is the 0-based position in the 20-slot
+feed — and post text is stored once — about a 150×
 reduction, which is why the whole corpus fits in a few hundred MB instead of 26 GB.
 
+## Trajectories: every activation, including the passive ones
+
+`actions.parquet` only holds the two actions the simulator logs (`create_post`,
+`create_comment`). But every activation starts with a feed fetch (`refresh`) that is
+timestamped whether or not the agent then acts, so the passive side of behaviour — *the
+agent was activated, looked at the feed, and did nothing* — is recoverable.
+`trajectories.parquet` has **one row per agent activation**:
+
+| column | meaning |
+|---|---|
+| `run_id`, `agent_id`, `model` | as elsewhere |
+| `activation_idx` | k-th activation of this agent in this run; `-1` is the step-0 intro post (written without a feed fetch) |
+| `step` | simulation tick, exact where aligned (see below), else NULL |
+| `refreshed_at` | wall-clock time the agent fetched its feed |
+| `status` | `ok` / `exhausted` (all retries failed to parse) / `exception`; NULL where unaligned |
+| `n_attempts` | generation attempts the harness made for this activation |
+| `action` | `create_post` / `create_comment` / `none` |
+| `acted_at`, `latency_s` | when the action landed, and seconds after the feed fetch |
+| `post_id` | own post (`create_post`) or **target** post (`create_comment`) |
+| `comment_id` | the comment written, if any |
+| `tool_name`, `tool_args` | the tool call the model actually emitted (JSON). Kept even when the action never landed — e.g. a comment aimed at a `post_id` that did not exist — so *chosen* and *realised* action can be told apart |
+| `response_text` | the model's visible output before the tool call: its stated deliberation, for families that state one |
+| `reasoning` | the model's hidden thinking channel, for families that expose one |
+| `finish_reason`, `errors` | generation finish reason; parse errors of failed attempts, `|`-joined |
+| `extra_in_window` | True for a 2nd+ action inside one feed-fetch window (16 runs; see below) |
+
+**How agents choose feed items.** Join `trajectories.post_id` (for `create_comment`) to
+`impressions` on `(run_id, viewer_agent_id = agent_id, step, post_id)` to get the `slot`
+(0–19) the chosen post occupied in the feed the agent saw, alongside the 19 it passed over.
+Then `response_text` / `reasoning` say *why*, where the family says so. Coverage of the
+stated deliberation is a property of the model, not of the harness:
+
+| model | activations | `response_text` | `reasoning` | passive |
+|---|---:|---:|---:|---:|
+| gpt-oss-20b | 247,489 | 0.4% | 4.8% | 0.3% |
+| Magistral-Small-2509 | 149,666 | 86.2% | 91.7% | 1.1% |
+| Qwen3-32B | 102,546 | 44.6% | 44.7% | 0.3% |
+| GLM-4-32B-0414 | 54,430 | 0.0% | 0.0% | 0.6% |
+| gemma-4-31B-it | 39,744 | 93.0% | 93.4% | 1.2% |
+
+gpt-oss and GLM-4 emit no visible deliberation at all; for those the decision record is the
+tool call plus the feed it was chosen from.
+
+**What is not kept.** The prompt itself. Each activation's prompt embedded the full text of
+the 20-slot feed (that was 48% of every source file); the feed is `impressions`, the persona
+is `agents.persona`, and the system directive is in the released code.
+
+**Alignment.** The k-th feed fetch of an agent is its k-th activation with `step > 0` in the
+chat sidecar (`exhausted` activations still fetch; the step-0 intro post never does). Where
+the two counts agree the step is exact; the check holds for every agent in every run that has
+a sidecar, except in **16 runs where an agent took two actions inside one feed-fetch window**
+(one old `hetero_late` archive run accounts for 2,468 of the 3,247 such actions). Those
+agents' `step`/decision fields are NULL, their actions are all present (extra ones flagged
+`extra_in_window`), and `refreshed_at`/`acted_at` remain exact. Overall `step` coverage in
+`trajectories` is **89.5% of activations**.
+
+Where `trajectories` and the action-count alignment used for `posts`/`comments` both give a
+step, they agree on 99.99% of rows; the disagreements are all inside those 16 runs.
+
 ## Known data-quality issues
 
 - **`OLMo-3.1-32B-Think` is degenerate**: 25.5% of its authored texts are under 15
@@ -124,6 +197,12 @@ reduction, which is why the whole corpus fits in a few hundred MB instead of 26
 - Runs missing a metadata sidecar (~1.6%) are excluded entirely; model attribution would be
   impossible.
 
+## Build
+
+The tables are produced by `build_divdata.py` (six core tables) then `build_trajectories.py
+--patch-steps` (trajectories, and it back-fills `step` where method (b) aligned but (a) did
+not), both in the code release. `BUILD_LOG.md` is the run log of the current build.
+
 ## Provenance
 
 `run_id` encodes the source path, so any row is traceable to the original database. Model

_trajectories_build_log.json

@@ -0,0 +1,45 @@
+{
+ "runs": 1296,
+ "no_chats": 41,
+ "agents": 28847,
+ "agents_aligned": 26786,
+ "activations": 748285,
+ "passive": 7066,
+ "windows_multi": 3247,
+ "agents_multi": 339,
+ "with_reasoning": 284788,
+ "with_response_text": 258949,
+ "with_tool_call": 642083,
+ "status": {
+  "ok": 670505,
+  "null": 75438,
+  "exhausted": 2263,
+  "exception": 79
+ },
+ "rows": 748285,
+ "size_mb": 169.6,
+ "step_coverage": 0.8992,
+ "action_counts": {
+  "create_comment": 652588,
+  "create_post": 88631,
+  "none": 7066
+ },
+ "patch_posts": {
+  "null_before": 15060,
+  "null_after": 8701,
+  "rows": 89171,
+  "agreement_on_overlap": 0.99995
+ },
+ "patch_comments": {
+  "null_before": 91744,
+  "null_after": 53403,
+  "rows": 652588,
+  "agreement_on_overlap": 0.99993
+ },
+ "patch_actions": {
+  "null_before": 106264,
+  "null_after": 61564,
+  "rows": 741219,
+  "agreement_on_overlap": 0.99993
+ }
+}

actions.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d5bb00e9517d03e843764db56ef7101ded854341e395928260ffac8f48ac858e
-size 6004815
+oid sha256:8f972f4cbabe672dfa8d5cb35abc1966ad2e27d227b1479ed68c53fcc102f1b1
+size 6070070

comments.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ab740177096feb1bf06ee4d98e401096d40e110e05d51efa681465f44f120227
-size 92243568
+oid sha256:e3b527acc88d0747be12821397069c82f13994dc51cea8cb704d23b2ea132b9f
+size 91803414

impressions.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7f9915f755462e2e2e3e568bbd0d60e973a9cfa186d56336b4a3fc7ac78c046f
-size 1553969
+oid sha256:b7e008d1d0c635dd95c30bc760ce6d4db9ac51db3d8a044e69a160578a014507
+size 1165781

posts.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9a6566f1f01359efd63658ee2e7a95844b54474f582899262040a41d1b711bbd
-size 23814484
+oid sha256:efc9e24dd7e1b9093bcb1237c3d382ff7d375f7a69825f3c6f560ca5235ad8ca
+size 23819999

trajectories.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:73828d61222e7de74783d7b989cc4e2fd29544f070435231ab22e0960389acfe
+size 169599974