[bug] Nondeterminism

[AlekseyKorshuk]

I have spotted some lines where the generator is used without the seed. Some of them might not be used, but the rest a critical for result consistency. Here is the list:

• concordia/examples/modular/environment/forbidden_fruit.py

  Line 112 in 9f317f5

  TREE = random.choice(FORBIDDEN_TREE_NAMES)

• concordia/examples/modular/environment/forbidden_fruit.py

  Line 897 in 9f317f5

  random.shuffle(main_player_configs)

• concordia/concordia/components/agent/person_representation.py

  Line 109 in 9f317f5

  self._names_detected = random.sample(self._names_detected,

• concordia/concordia/components/game_master/conversation.py

  Line 188 in 9f317f5

  state=random.choice(CONVERSATIONALIST_STYLES),

• concordia/concordia/contrib/components/game_master/daily_activities.py

  Line 399 in 9f317f5

  random.shuffle(events)

• concordia/concordia/contrib/components/game_master/world_background_and_relevance.py

  Line 122 in 9f317f5

  permuted_world_building_prompt_questions = random.sample(

• concordia/concordia/contrib/components/game_master/world_background_and_relevance.py

  Line 155 in 9f317f5

  example_elements = random.sample(elements, 3)

• concordia/concordia/document/interactive_document.py

  Line 274 in 9f317f5

  response = random.choice(candidates)

• concordia/concordia/thought_chains/thought_chains.py

  Line 366 in 9f317f5

  random.shuffle(inactive_players_who_acted)

• concordia/examples/modular/environment/modules/player_traits_and_styles.py

  Line 191 in 9f317f5

  return random.choice(FLOWERY_TRAITS)

• concordia/examples/modular/environment/modules/player_traits_and_styles.py

  Line 193 in 9f317f5

  return random.choice(TRAITS)

• concordia/examples/modular/environment/modules/player_traits_and_styles.py

  Line 206 in 9f317f5

  return random.choice(CONVERSATION_STYLES).format(player_name=player_name)

• concordia/examples/modular/environment/modules/pub_coordination_capetown.py

  Line 200 in 9f317f5

  pubs = random.sample(list(PUB_PREFERENCES.keys()), NUM_PUBS)

• concordia/examples/modular/environment/modules/pub_coordination_london.py

  Line 283 in 9f317f5

  pubs = random.sample(list(PUB_PREFERENCES.keys()), NUM_PUBS)

• concordia/examples/open_ended/open_ended.py

  Line 178 in 9f317f5

  random.shuffle(main_player_configs)

[jzleibo]

Here's how we're thinking about this. We could make the lines you mention use the seed, and we should and will do that in the long run. However, this won't actually make the result deterministic. The problem is that these are all places that will be called after substantial use of the LLM API has already happened. But the LLM API could be delayed for network traffic reasons, rate limits, etc, any number of things. These would all end up advancing the seed. So just making the lines you mention take a seed parameter would not actually yield a deterministic result in practice.

What we have done to take as big a step as possible toward determinism is to make sure that all the random choices that happen in environment setup, before we use the LLM, are all seeded properly. This part of the variation really can be made deterministic, so we have done that.