Reality Monitoring Task with Feedback — ps-parser branch¶

Runs the two-phase RM word-pair task with trial-level corrective feedback using the ps-parser API.

Task structure¶

Phase	Trials	Parser	Response fields
Encoding	`imagined_N`, `perceived_N`	`Response_part_1_rm`	`Word_2 (str)`, `Rating (0–100)`
Test	`test:imagined_N`, `test:perceived_N`	`Response_part_2_rm`	`Judgment (internal/external)`, `Confidence (1–6)`

What is new vs the old archive version¶

	Old archive (`parser='dynamic'`)	ps-parser branch
Parser routing	Hardcoded conditional edges in LangGraph	Per-trial `"parser"` key in task JSON
Card-level parser	`card_in.parser = 'dynamic'`	`card_in.parser = '0'` (per-trial JSON handles it)
Feedback bool	`feedback = '1'`	`feedback = True`
Feedback handler	Re-instantiated every trial	Instantiated once per simulation
Cross-trial word list	Module-level global (breaks with N>1)	`self.all_words_in_use` — per-instance

Model: Groq — openai/gpt-oss-120b

1. Setup¶

In [1]:

Copied!





import os, sys, ast, json, nltk
from pathlib import Path
from dotenv import load_dotenv

nltk.download('words', quiet=True)
load_dotenv()

# ── LLM ─────────────────────────────────────────────────────────────────────
# psychscanner automatically loads GROQ_API_KEY from the environment
MODEL_NAME   = 'openai/gpt-oss-120b'
MODEL_FAMILY = 'groq'

# ── Paths ─────────────────────────────────────────────────────────────────
EXAMPLES_DIR = Path(os.getcwd())   # notebook lives in examples/
# Update TASKS_DIR to point to your local RM task files
TASKS_DIR    = Path(os.getenv('RM_TASKS_DIR', Path.cwd() / 'tasks'))
RUN_DIR      = EXAMPLES_DIR / '_rm_fb_tutorial_runs'

task_files = {
    'demo_5t'     : TASKS_DIR / 'rm_2op_convo_5t_feedback_allparts.json',
    'demo_5t_rev' : TASKS_DIR / 'rm_2op_convo_5t_feedback_allparts_r.json',
    'full'        : TASKS_DIR / 'rm_2op_convo_feedback_allparts.json',
    'full_rev'    : TASKS_DIR / 'rm_2op_convo_feedback_allparts_r.json',
}
for name, p in task_files.items():
    print(f'  {name:12} -> {p.name}  exists={p.exists()}')
import os, sys, ast, json, nltk
from pathlib import Path
from dotenv import load_dotenv

nltk.download('words', quiet=True)
load_dotenv()

# ── LLM ─────────────────────────────────────────────────────────────────────
# psychscanner automatically loads GROQ_API_KEY from the environment
MODEL_NAME   = 'openai/gpt-oss-120b'
MODEL_FAMILY = 'groq'

# ── Paths ─────────────────────────────────────────────────────────────────
EXAMPLES_DIR = Path(os.getcwd())   # notebook lives in examples/
# Update TASKS_DIR to point to your local RM task files
TASKS_DIR    = Path(os.getenv('RM_TASKS_DIR', Path.cwd() / 'tasks'))
RUN_DIR      = EXAMPLES_DIR / '_rm_fb_tutorial_runs'

task_files = {
    'demo_5t'     : TASKS_DIR / 'rm_2op_convo_5t_feedback_allparts.json',
    'demo_5t_rev' : TASKS_DIR / 'rm_2op_convo_5t_feedback_allparts_r.json',
    'full'        : TASKS_DIR / 'rm_2op_convo_feedback_allparts.json',
    'full_rev'    : TASKS_DIR / 'rm_2op_convo_feedback_allparts_r.json',
}
for name, p in task_files.items():
    print(f'  {name:12} -> {p.name}  exists={p.exists()}')

  demo_5t      -> rm_2op_convo_5t_feedback_allparts.json  exists=True
  demo_5t_rev  -> rm_2op_convo_5t_feedback_allparts_r.json  exists=True
  full         -> rm_2op_convo_feedback_allparts.json  exists=True
  full_rev     -> rm_2op_convo_feedback_allparts_r.json  exists=True

In [2]:

Copied!





import psychscanner as psy
from psychscanner import ExpCard, ExpCardInit, ScannerModel
from psychscanner.parsers import Response_part_1_rm, Response_part_2_rm

if str(EXAMPLES_DIR) not in sys.path:
    sys.path.insert(0, str(EXAMPLES_DIR))
from rm_msg_injection import Stim_Trial_Injection

print('psychscanner version:', psy.__version__)
print('Response_part_1_rm fields:', list(Response_part_1_rm.model_fields.keys()))
print('Response_part_2_rm fields:', list(Response_part_2_rm.model_fields.keys()))
import psychscanner as psy
from psychscanner import ExpCard, ExpCardInit, ScannerModel
from psychscanner.parsers import Response_part_1_rm, Response_part_2_rm

if str(EXAMPLES_DIR) not in sys.path:
    sys.path.insert(0, str(EXAMPLES_DIR))
from rm_msg_injection import Stim_Trial_Injection

print('psychscanner version:', psy.__version__)
print('Response_part_1_rm fields:', list(Response_part_1_rm.model_fields.keys()))
print('Response_part_2_rm fields:', list(Response_part_2_rm.model_fields.keys()))

psychscanner version: 0.1.0
Response_part_1_rm fields: ['Word_2', 'Rating']
Response_part_2_rm fields: ['Judgment', 'Confidence']

/opt/anaconda3/envs/psyscan/lib/python3.11/site-packages/langgraph/checkpoint/base/__init__.py:17: LangChainPendingDeprecationWarning: The default value of `allowed_objects` will change in a future version. Pass an explicit value (e.g., allowed_objects='messages' or allowed_objects='core') to suppress this warning.
  from langgraph.checkpoint.serde.jsonplus import JsonPlusSerializer

2. Per-trial parser routing in the task JSON¶

The old archive used parser='dynamic', which hardcoded two LangGraph nodes (runnable_resp1node / runnable_resp2node) and routed between them based on trcode. That sentinel no longer exists in ps-parser.

The task JSON files have been updated: every trial now carries a "parser" key that names the registered parser class to use for that trial. The ps-parser branch reads this key at invocation time and resolves the class from the PARSER_REGISTRY — no hardcoding needed.

Encoding trial:  "parser": "Response_part_1_rm"   →  Word_2, Rating
Test trial:      "parser": "Response_part_2_rm"   →  Judgment, Confidence

In [3]:

Copied!





task_json = json.loads(task_files['demo_5t'].read_text())

print('Task name  :', task_json['taskname'])
print('Chain type :', task_json['chain_type'])
print('Top-level parser (fallback):', task_json['parser'])   # '0' = none
print()

items = task_json['items']
print(f'{"trcode":<24} {"trial parser":<22} {"stimulus / corrAns"}')
print('-' * 80)
for trcode, trials in items.items():
    t = trials[0]
    stim = t.get('stimulus', {})
    detail = stim.get('Word_Pair', stim.get('Test_Word', stim))
    corr = t.get('corrAns', '')
    extra = f'corrAns={corr}' if corr else str(detail)[:45]
    print(f"{trcode:<24} {t.get('parser','(none)'):<22} {extra}")
task_json = json.loads(task_files['demo_5t'].read_text())

print('Task name  :', task_json['taskname'])
print('Chain type :', task_json['chain_type'])
print('Top-level parser (fallback):', task_json['parser'])   # '0' = none
print()

items = task_json['items']
print(f'{"trcode":<24} {"trial parser":<22} {"stimulus / corrAns"}')
print('-' * 80)
for trcode, trials in items.items():
    t = trials[0]
    stim = t.get('stimulus', {})
    detail = stim.get('Word_Pair', stim.get('Test_Word', stim))
    corr = t.get('corrAns', '')
    extra = f'corrAns={corr}' if corr else str(detail)[:45]
    print(f"{trcode:<24} {t.get('parser','(none)'):<22} {extra}")

Task name  : rm_2op_5t_convo_fb
Chain type : task
Top-level parser (fallback): 0

trcode                   trial parser           stimulus / corrAns
--------------------------------------------------------------------------------
imagined_1               Response_part_1_rm     {'word_1': 'sandpaper', 'word_2': '________'}
perceived_2              Response_part_1_rm     {'word_1': 'sugar', 'word_2': 'sweet'}
imagined_3               Response_part_1_rm     {'word_1': 'suburb', 'word_2': '________'}
perceived_4              Response_part_1_rm     {'word_1': 'chair', 'word_2': 'cushion'}
perceived_5              Response_part_1_rm     {'word_1': 'letter', 'word_2': 'tip'}
imagined_6               Response_part_1_rm     {'word_1': 'building', 'word_2': '________'}
imagined_7               Response_part_1_rm     {'word_1': 'blouse', 'word_2': '________'}
perceived_8              Response_part_1_rm     {'word_1': 'pin', 'word_2': 'haystack'}
imagined_9               Response_part_1_rm     {'word_1': 'fire', 'word_2': '________'}
perceived_11             Response_part_1_rm     {'word_1': 'sky', 'word_2': 'airplane'}
test:imagined_1          Response_part_2_rm     corrAns=internal
test:perceived_2         Response_part_2_rm     corrAns=external
test:imagined_3          Response_part_2_rm     corrAns=internal
test:perceived_4         Response_part_2_rm     corrAns=external
test:perceived_5         Response_part_2_rm     corrAns=external
test:imagined_6          Response_part_2_rm     corrAns=internal
test:imagined_7          Response_part_2_rm     corrAns=internal
test:perceived_8         Response_part_2_rm     corrAns=external
test:imagined_9          Response_part_2_rm     corrAns=internal
test:perceived_11        Response_part_2_rm     corrAns=external

3. Feedback handler — `Stim_Trial_Injection`¶

rm_msg_injection.py provides Stim_Trial_Injection(FeedbackBase). Because ps-parser instantiates it once per simulation, self.all_words_in_use accumulates words across all encoding trials for that participant.

on_response receives a plain dict already parsed by TaskRunner:

Encoding: {'Word_2': '...', 'Rating': 75.0}
Test: {'Judgment': 'internal', 'Confidence': 5}

In [4]:

Copied!





from rm_msg_injection import generate_fb_response

# ── Imagined encoding trial — correct imagined word ───────────────────────────
trial_enc = {
    'trcode'   : 'imagined_1',
    'stimulus' : {'Word_Pair': {'word_1': 'sandpaper', 'word_2': '________'}},
}
words_seen = []
fb_enc = generate_fb_response(trial_enc, {'Word_2': 'velvet', 'Rating': 60.0}, words_seen)
print('Encoding feedback (imagined, correct):')
print(json.dumps(json.loads(fb_enc), indent=2))
print('words tracked:', words_seen)
from rm_msg_injection import generate_fb_response

# ── Imagined encoding trial — correct imagined word ───────────────────────────
trial_enc = {
    'trcode'   : 'imagined_1',
    'stimulus' : {'Word_Pair': {'word_1': 'sandpaper', 'word_2': '________'}},
}
words_seen = []
fb_enc = generate_fb_response(trial_enc, {'Word_2': 'velvet', 'Rating': 60.0}, words_seen)
print('Encoding feedback (imagined, correct):')
print(json.dumps(json.loads(fb_enc), indent=2))
print('words tracked:', words_seen)

Encoding feedback (imagined, correct):
{
  "Feedback on previous response": {
    "response feedback": "**CORRECT: Last trial was an imagined trial type. Your 'Word_2' correctly followed the given instructions.**",
    "rating feedback": "**CORRECT: Your 'rating' value is in the instructed rating scale range.**",
    "overall feedback": "**CORRECT Answer!** Well answered, good job!"
  }
}
words tracked: ['sandpaper', 'velvet']

In [5]:

Copied!





# ── Test trial — wrong judgment ───────────────────────────────────────────────
trial_test = {
    'trcode'  : 'test:imagined_1',
    'stimulus': {'Test_Word': 'sandpaper'},
    'corrAns' : 'internal',
}
fb_test = generate_fb_response(trial_test, {'Judgment': 'external', 'Confidence': 3}, [])
print('Test feedback (wrong judgment):')
print(json.dumps(json.loads(fb_test), indent=2))
# ── Test trial — wrong judgment ───────────────────────────────────────────────
trial_test = {
    'trcode'  : 'test:imagined_1',
    'stimulus': {'Test_Word': 'sandpaper'},
    'corrAns' : 'internal',
}
fb_test = generate_fb_response(trial_test, {'Judgment': 'external', 'Confidence': 3}, [])
print('Test feedback (wrong judgment):')
print(json.dumps(json.loads(fb_test), indent=2))

Test feedback (wrong judgment):
{
  "Feedback on previous response": {
    "response feedback": "**INCORRECT: The correct 'Judgment' was 'internal'. It might also be incorrect due to poor formatting. Carefully follow all trial instructions about the source-judgment options.**",
    "rating feedback": "**CORRECT: Your 'Confidence' value is in the instructed rating scale range (1\u20136).**",
    "overall feedback": "**INCORRECT Answer!** Follow all trial instructions more accurately."
  }
}

4. Build the experiment card¶

The key difference from the old archive script:

In [6]:

Copied!





# Old archive API vs ps-parser API
print('=== Card-level differences: old archive vs ps-parser ===')
print()
print('OLD (archive):')
print("  card_in.parser   = 'dynamic'   # hardcoded trcode routing in LangGraph")
print("  card_in.feedback = '1'         # string")
print()
print('NEW (ps-parser):')
print("  card_in.parser   = '0'         # no card-level parser; per-trial JSON 'parser' field used")
print("  card_in.feedback = True        # bool  ('1' still works for backward compat)")
# Old archive API vs ps-parser API
print('=== Card-level differences: old archive vs ps-parser ===')
print()
print('OLD (archive):')
print("  card_in.parser   = 'dynamic'   # hardcoded trcode routing in LangGraph")
print("  card_in.feedback = '1'         # string")
print()
print('NEW (ps-parser):')
print("  card_in.parser   = '0'         # no card-level parser; per-trial JSON 'parser' field used")
print("  card_in.feedback = True        # bool  ('1' still works for backward compat)")

=== Card-level differences: old archive vs ps-parser ===

OLD (archive):
  card_in.parser   = 'dynamic'   # hardcoded trcode routing in LangGraph
  card_in.feedback = '1'         # string

NEW (ps-parser):
  card_in.parser   = '0'         # no card-level parser; per-trial JSON 'parser' field used
  card_in.feedback = True        # bool  ('1' still works for backward compat)

In [7]:

Copied!





card_in = ExpCardInit()

# ── Model ─────────────────────────────────────────────────────────────────────
card_in.model      = MODEL_NAME
card_in.family     = MODEL_FAMILY
card_in.parameters = {'temperature': 0}

# ── Task ─────────────────────────────────────────────────────────────────────
card_in.task_file  = task_files['demo_5t']  # 10 encoding + 10 test trials
card_in.parser     = '0'   # no card-level default — per-trial JSON 'parser' field handles routing
                            # encoding → Response_part_1_rm, test → Response_part_2_rm

# ── Memory ───────────────────────────────────────────────────────────────────
card_in.memory     = 'Convo'   # full conversation across all trials
card_in.chain_type = 'task'    # single thread per participant

# ── Participant ───────────────────────────────────────────────────────────────
card_in.cogtype    = 'no'
card_in.nsim       = 1         # 1 simulated participant; set to 100 for a full study

# ── Feedback ─────────────────────────────────────────────────────────────────
card_in.feedback    = True                  # bool (was '1' in the old archive)
card_in.feedback_fn = Stim_Trial_Injection  # class — instantiated once per simulation

# ── Output ───────────────────────────────────────────────────────────────────
card_in.proj_dir      = RUN_DIR
card_in.projectname   = 'RM_FB_TUTORIAL'
card_in.tunnel_status = '0'

expcard = ExpCard(card_in)
print('card-level parser resolved to:', expcard.parser)  # None — per-trial JSON overrides
card_in = ExpCardInit()

# ── Model ─────────────────────────────────────────────────────────────────────
card_in.model      = MODEL_NAME
card_in.family     = MODEL_FAMILY
card_in.parameters = {'temperature': 0}

# ── Task ─────────────────────────────────────────────────────────────────────
card_in.task_file  = task_files['demo_5t']  # 10 encoding + 10 test trials
card_in.parser     = '0'   # no card-level default — per-trial JSON 'parser' field handles routing
                            # encoding → Response_part_1_rm, test → Response_part_2_rm

# ── Memory ───────────────────────────────────────────────────────────────────
card_in.memory     = 'Convo'   # full conversation across all trials
card_in.chain_type = 'task'    # single thread per participant

# ── Participant ───────────────────────────────────────────────────────────────
card_in.cogtype    = 'no'
card_in.nsim       = 1         # 1 simulated participant; set to 100 for a full study

# ── Feedback ─────────────────────────────────────────────────────────────────
card_in.feedback    = True                  # bool (was '1' in the old archive)
card_in.feedback_fn = Stim_Trial_Injection  # class — instantiated once per simulation

# ── Output ───────────────────────────────────────────────────────────────────
card_in.proj_dir      = RUN_DIR
card_in.projectname   = 'RM_FB_TUTORIAL'
card_in.tunnel_status = '0'

expcard = ExpCard(card_in)
print('card-level parser resolved to:', expcard.parser)  # None — per-trial JSON overrides

----<PROJECT AND DATA ROOT DIRECTORY>----
	Project root dir: /Users/saurabhext/Documents/PSYCHSCANNER/psychscanner/examples/_rm_fb_tutorial_runs
	Simulation data root dir: /Users/saurabhext/Documents/PSYCHSCANNER/psychscanner/examples/_rm_fb_tutorial_runs/RM_FB_TUTORIAL/rm_2op_5t_convo_fb/groq_openai/gpt-oss-120b_Convo
----<>----
card-level parser resolved to: None

5. Run¶

In [8]:

Copied!

scanner    = ScannerModel(expcard=expcard)
simulation = scanner.run()
print(f'\nDone — {len(simulation)} participant(s), {len(simulation[0])} trials each.')
scanner    = ScannerModel(expcard=expcard)
simulation = scanner.run()
print(f'\nDone — {len(simulation)} participant(s), {len(simulation[0])} trials each.')

--<chat model>-- profile={'max_input_tokens': 131072, 'max_output_tokens': 32768, 'image_inputs': False, 'audio_inputs': False, 'video_inputs': False, 'image_outputs': False, 'audio_outputs': False, 'video_outputs': False, 'reasoning_output': True, 'tool_calling': True} client=<groq.resources.chat.completions.Completions object at 0x152c003d0> async_client=<groq.resources.chat.completions.AsyncCompletions object at 0x151eb1310> model_name='openai/gpt-oss-120b' temperature=1e-08 model_kwargs={} groq_api_key=SecretStr('**********')

2026-05-07 02:12:51.631 | CRITICAL | psychscanner.session_tunnel.session_tunnel:create_tunnel:152 - BEGIN

TOTAL RUNS: 1	RESUME IDX: None
----<>---- task running

20it [04:02, 12.14s/it]
2026-05-07 02:16:54.436 | INFO     | psychscanner.session_tunnel.session_tunnel:scan_checkpoint:184 - scan-checkpoint

----<scanned runs>---- i = 0

2026-05-07 02:16:54.441 | CRITICAL | psychscanner.session_tunnel.session_tunnel:end_checkpoint:163 - END

Done — 1 participant(s), 20 trials each.

6. Results¶

In [9]:

Copied!





def decode(pred_resp):
    content = pred_resp.content if hasattr(pred_resp, 'content') else str(pred_resp)
    try:
        return ast.literal_eval(content)
    except Exception:
        return {'_raw': content[:80]}

trials = simulation[0]

# ── Encoding phase ────────────────────────────────────────────────────────────
enc_trials  = [t for t in trials if 'test' not in t['trcode']]
test_trials = [t for t in trials if 'test'     in t['trcode']]

print('ENCODING PHASE')
print(f'{"trcode":<20} {"Word_2":<16} {"Rating":<8} {"feedback sent?"}')
print('-' * 65)
for t in enc_trials:
    r = decode(t['pred_resp'])
    print(f"{t['trcode']:<20} {str(r.get('Word_2','?')):<16} "
          f"{str(r.get('Rating','?')):<8} {'yes' if t.get('fb_response') else 'no'}")

print()
print('TEST PHASE')
print(f'{"trcode":<24} {"Judgment":<12} {"corrAns":<12} {"Conf":<6} {"correct?"}')
print('-' * 65)
correct = 0
for t in test_trials:
    r = decode(t['pred_resp'])
    j = str(r.get('Judgment', '?')).lower()
    c = str(t.get('corrAns', '')).lower()
    ok = j == c
    if ok:
        correct += 1
    print(f"{t['trcode']:<24} {j:<12} {c:<12} {str(r.get('Confidence','?')):<6} {'✓' if ok else '✗'}")

print()
print(f'Source-monitoring accuracy: {correct}/{len(test_trials)} = {correct/len(test_trials):.0%}')
def decode(pred_resp):
    content = pred_resp.content if hasattr(pred_resp, 'content') else str(pred_resp)
    try:
        return ast.literal_eval(content)
    except Exception:
        return {'_raw': content[:80]}

trials = simulation[0]

# ── Encoding phase ────────────────────────────────────────────────────────────
enc_trials  = [t for t in trials if 'test' not in t['trcode']]
test_trials = [t for t in trials if 'test'     in t['trcode']]

print('ENCODING PHASE')
print(f'{"trcode":<20} {"Word_2":<16} {"Rating":<8} {"feedback sent?"}')
print('-' * 65)
for t in enc_trials:
    r = decode(t['pred_resp'])
    print(f"{t['trcode']:<20} {str(r.get('Word_2','?')):<16} "
          f"{str(r.get('Rating','?')):<8} {'yes' if t.get('fb_response') else 'no'}")

print()
print('TEST PHASE')
print(f'{"trcode":<24} {"Judgment":<12} {"corrAns":<12} {"Conf":<6} {"correct?"}')
print('-' * 65)
correct = 0
for t in test_trials:
    r = decode(t['pred_resp'])
    j = str(r.get('Judgment', '?')).lower()
    c = str(t.get('corrAns', '')).lower()
    ok = j == c
    if ok:
        correct += 1
    print(f"{t['trcode']:<24} {j:<12} {c:<12} {str(r.get('Confidence','?')):<6} {'✓' if ok else '✗'}")

print()
print(f'Source-monitoring accuracy: {correct}/{len(test_trials)} = {correct/len(test_trials):.0%}')

ENCODING PHASE
trcode               Word_2           Rating   feedback sent?
-----------------------------------------------------------------
imagined_1           abrasive         85.0     yes
perceived_2          sweet            92.0     yes
imagined_3           neighborhood     90.0     yes
perceived_4          cushion          88.0     yes
perceived_5          tip              30.0     yes
imagined_6           construction     85.0     yes
imagined_7           shirt            80.0     yes
perceived_8          haystack         25.0     yes
imagined_9           smoke            92.0     yes
perceived_11         airplane         85.0     yes

TEST PHASE
trcode                   Judgment     corrAns      Conf   correct?
-----------------------------------------------------------------
test:imagined_1          internal     internal     6      ✓
test:perceived_2         external     external     6      ✓
test:imagined_3          internal     internal     6      ✓
test:perceived_4         external     external     6      ✓
test:perceived_5         external     external     6      ✓
test:imagined_6          internal     internal     6      ✓
test:imagined_7          internal     internal     6      ✓
test:perceived_8         external     external     6      ✓
test:imagined_9          internal     internal     6      ✓
test:perceived_11        external     external     6      ✓

Source-monitoring accuracy: 10/10 = 100%

In [10]:

Copied!





# Show the full feedback JSON injected after the first encoding trial
t = enc_trials[0]
print(f"Trial: {t['trcode']}")
print(f"Response: {decode(t['pred_resp'])}")
print()
if t['fb_response']:
    print('Feedback injected before next trial:')
    try:
        print(json.dumps(json.loads(t['fb_response']), indent=2))
    except Exception:
        print(t['fb_response'])
else:
    print('No feedback for this trial.')
# Show the full feedback JSON injected after the first encoding trial
t = enc_trials[0]
print(f"Trial: {t['trcode']}")
print(f"Response: {decode(t['pred_resp'])}")
print()
if t['fb_response']:
    print('Feedback injected before next trial:')
    try:
        print(json.dumps(json.loads(t['fb_response']), indent=2))
    except Exception:
        print(t['fb_response'])
else:
    print('No feedback for this trial.')

Trial: imagined_1
Response: {'Word_2': 'abrasive', 'Rating': 85.0}

Feedback injected before next trial:
{
  "Feedback on previous response": {
    "response feedback": "**CORRECT: Last trial was an imagined trial type. Your 'Word_2' correctly followed the given instructions.**",
    "rating feedback": "**CORRECT: Your 'rating' value is in the instructed rating scale range.**",
    "overall feedback": "**CORRECT Answer!** Well answered, good job!"
  }
}

7. Using the callable form instead of per-trial JSON (alternative)¶

If you prefer to keep the routing logic in Python rather than in the task JSON, you can use a Form B callable at the card level and leave the task JSON trial "parser" field absent.

In [11]:

Copied!





from psychscanner.parsers import Response_part_1_rm, Response_part_2_rm

# Form B: callable at card level — routes by trcode
rm_parser_dispatch = lambda trcode: (
    Response_part_2_rm if 'test' in trcode else Response_part_1_rm
)

# Verify routing
print('encode trial  ->', rm_parser_dispatch('imagined_1'))
print('test trial    ->', rm_parser_dispatch('test:imagined_1'))
print()
print('To use: set card_in.parser = rm_parser_dispatch')
print('Per-trial JSON "parser" (Form A) still takes priority over this callable.')
from psychscanner.parsers import Response_part_1_rm, Response_part_2_rm

# Form B: callable at card level — routes by trcode
rm_parser_dispatch = lambda trcode: (
    Response_part_2_rm if 'test' in trcode else Response_part_1_rm
)

# Verify routing
print('encode trial  ->', rm_parser_dispatch('imagined_1'))
print('test trial    ->', rm_parser_dispatch('test:imagined_1'))
print()
print('To use: set card_in.parser = rm_parser_dispatch')
print('Per-trial JSON "parser" (Form A) still takes priority over this callable.')

encode trial  -> <class 'psychscanner.datasets.prompts.parser_tasks.Response_part_1_rm'>
test trial    -> <class 'psychscanner.datasets.prompts.parser_tasks.Response_part_2_rm'>

To use: set card_in.parser = rm_parser_dispatch
Per-trial JSON "parser" (Form A) still takes priority over this callable.

8. Scaling to many participants¶

Change nsim and optionally switch to the full 20-pair task or the reversed counterbalance list. Each participant gets its own Stim_Trial_Injection instance so word lists never leak between them.

In [12]:

Copied!





# Uncomment to run a full multi-participant study
#
# card_in.nsim         = 100
# card_in.task_file    = task_files['full']        # 20 encoding + 20 test trials
# # or reversed counterbalance list:
# # card_in.task_file  = task_files['full_rev']
# card_in.tunnel_status = '1'                      # enable checkpointing for long runs
#
# expcard_full = ExpCard(card_in)
# scanner_full = ScannerModel(expcard=expcard_full)
# simulation_full = scanner_full.run()
print('Scaling config ready — uncomment above to run.')
# Uncomment to run a full multi-participant study
#
# card_in.nsim         = 100
# card_in.task_file    = task_files['full']        # 20 encoding + 20 test trials
# # or reversed counterbalance list:
# # card_in.task_file  = task_files['full_rev']
# card_in.tunnel_status = '1'                      # enable checkpointing for long runs
#
# expcard_full = ExpCard(card_in)
# scanner_full = ScannerModel(expcard=expcard_full)
# simulation_full = scanner_full.run()
print('Scaling config ready — uncomment above to run.')

Scaling config ready — uncomment above to run.