Is there a way to pipe the answer of an agent to a question, inside the text of another question?

[andreifoldes]

Qualtrics has a useful feature when the answer of a participant can be piped into subsequent question text.
Can this be done with EDSL?

q_mostRecentGame = QuestionFreeText(
    question_name="mostRecentGame",
    question_text="What is the name of the game you most recently played?\n\nThis text will be copied into certain other questions, so please write only the game name and use abbreviations if needed."
)

q_bangs_1 = QuestionLinearScale(
    question_name="bangs_1",
    question_text="In my most recent session of {{mostRecentGame}}... I could play in the way I wanted.",
    question_options=[1, 2, 3, 4, 5, 6, 7],
    option_labels={1: "Strongly Disagree", 7: "Strongly Agree"}
)

[johnjosephhorton]

It can! https://docs.expectedparrot.com/en/latest/surveys.html#id2
cc: @rbyh

[andreifoldes]

Amazing, thank you I missed it -.-

[andreifoldes]

I think I'm doing something wrong with it... as I'm getting "jinja2.exceptions.UndefinedError: 'mostRecentGame' is undefined"

Current implementation

q_mostRecentGame = QuestionFreeText(
    question_name="mostRecentGame",
    question_text="What is the name of the game you most recently played?\n\nThis text will be copied into certain other questions, so please write only the game name and use abbreviations if needed."
)

q_bangs_1 = QuestionLinearScale(
    question_name="bangs_1",
    question_text="In my most recent session of {{ mostRecentGame.answer }}... I could play in the way I wanted.",
    question_options=[1, 2, 3, 4, 5, 6, 7],
    option_labels={1: "Strongly Disagree", 7: "Strongly Agree"}
)

survey = Survey(questions = [
    q_played24hr,
    q_socialGaming,
    q_mostRecentGame,
    q_bangs_1,
    q_bangs_2,
    q_bangs_3,
    q_bangs_4,
    q_bangs_5,
    q_bangs_6,
    q_displacedActivity,
    q_lifeSat,
    q_affectiveValence,
    q_bpnsfs_1,
    q_bpnsfs_2,
    q_bpnsfs_3,
    q_bpnsfs_4,
    q_bpnsfs_5,
    q_bpnsfs_6,
    q_sd_0,
    q_sd_1_hour,
    q_sd_1_minute,
    q_sd_1_am_pm,
    q_sd_2_hour,
    q_sd_2_minute,
    q_sd_2_am_pm,
    q_sd_3,
    q_sd_4_hour,
    q_sd_4_minute,
    q_sd_4_am_pm,
    q_sd_5_hour,
    q_sd_5_minute,
    q_sd_5_am_pm,
    q_sd_6,
    #q_timeUse,
    q_stress_argument,
    q_howStress_arg,
    q_stress_letpass,
    q_howStress_letPass,
    q_stress_school,
    q_howStress_school,
    q_stress_home,
    q_howStress_home,
    q_stress_discrim,
    q_howStress_discrim,
    q_stress_relative,
    q_howStress_relative,
    q_stress_other,
    q_howStress_other
])

from edsl import Model

Model.available()
model = Model("gpt-4o-mini")
model.parameters['temperature'] = 0.7


#%%
survey_full_1= survey.set_lagged_memory(6)
results = survey_full_1.by(agents_list).by(model).run(progress_bar=True)

Traceback (most recent call last):
  File "C:\Users\saptaf1\AppData\Local\miniconda3\Lib\site-packages\edsl\jobs\tasks\QuestionTaskCreator.py", line 234, in _run_task_async
    return await self._run_focal_task(debug)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\saptaf1\AppData\Local\miniconda3\Lib\site-packages\edsl\jobs\tasks\QuestionTaskCreator.py", line 125, in _run_focal_task
    requested_tokens = self.estimated_tokens()
                       ^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\saptaf1\AppData\Local\miniconda3\Lib\site-packages\edsl\jobs\tasks\QuestionTaskCreator.py", line 100, in estimated_tokens
    return self.token_estimator(self.question)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\saptaf1\AppData\Local\miniconda3\Lib\site-packages\edsl\jobs\interviews\InterviewTaskBuildingMixin.py", line 160, in _get_estimated_request_tokens
    for prompt in invigilator.get_prompts().values():
                  ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\saptaf1\AppData\Local\miniconda3\Lib\site-packages\edsl\agents\PromptConstructionMixin.py", line 372, in get_prompts
    prior_question_memory=self.prior_question_memory_prompt,
                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\saptaf1\AppData\Local\miniconda3\Lib\site-packages\edsl\agents\PromptConstructionMixin.py", line 329, in prior_question_memory_prompt
    ).render(self.scenario)
      ^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\saptaf1\AppData\Local\miniconda3\Lib\site-packages\edsl\prompts\Prompt.py", line 208, in render
    new_text = self._render(
               ^^^^^^^^^^^^^
  File "C:\Users\saptaf1\AppData\Local\miniconda3\Lib\site-packages\edsl\prompts\Prompt.py", line 236, in _render
    rendered_text = env.from_string(text).render(
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\saptaf1\AppData\Local\miniconda3\Lib\site-packages\jinja2\environment.py", line 1304, in render
    self.environment.handle_exception()
  File "C:\Users\saptaf1\AppData\Local\miniconda3\Lib\site-packages\jinja2\environment.py", line 939, in handle_exception
    raise rewrite_traceback_stack(source=source)
  File "<template>", line 14, in top-level template code
  File "C:\Users\saptaf1\AppData\Local\miniconda3\Lib\site-packages\jinja2\environment.py", line 487, in getattr
    return getattr(obj, attribute)
           ^^^^^^^^^^^^^^^^^^^^^^^
jinja2.exceptions.UndefinedError: 'mostRecentGame' is undefined

[rbyh]

I am not able to reproduce this error with this code but I do not have the agents -- could you please share those details?

[rbyh]

Notebook I'm running to try to reproduce: https://www.expectedparrot.com/content/6e0d1624-1062-4890-86c8-7b13628771ed

[rbyh]

Ok I have reproduced the error (notebook is updated) -- it is a bug that occurs where there are questions after the questions that takes a piped answer:

This works fine:

survey = Survey(questions = [
    q_mostRecentGame,
    q_bangs_1 # has piped answer for mostRecentGame
])

This will generate the error:

survey = Survey(questions = [
    q_mostRecentGame,
    q_bangs_1,
    QuestionFreeText.example()
])

[andreifoldes]

Nice catch!!

[rbyh]

THANK YOU @andreifoldes !

[johnjosephhorton]

I will get started on a fix ASAP - thanks for reporting @andreifoldes !

[johnjosephhorton]

Ok, candidate fix is here: #899
Apologies & thanks for flagging it.
If it passes, we'll merge into main & do a new release to pypi cc: @apostolosfilippas