Actual Lecture 6

What is the formula for Bayes' rule?

In the context of Bayesian updates, what does Bayes' rule describe?

In Bayesian models of communication, reasoning moves beyond the plain message to consider what has been said, why it was said, and also what _____.

In the Matthias recommendation letter example, why is mentioning that he 'can indent his programming code very nicely' a negative signal?

What general communication principle dictates that a speaker should use an expression that is true and more informative than other true alternatives?

According to Grice's Maxim of _____, speakers should be maximally informative given the purpose of the conversation, but not redundantly so.

What is the term for the pragmatic inference that arises when a speaker's choice of a less specific term implies the negation of a more specific, salient alternative?

In the example 'I hurt my finger,' why do listeners infer 'not the thumb'?

What is the name for the recent attempt to combine the Gricean approach to pragmatics with probabilistic models to explain complex empirical data?

What experimental tool is used to study the nature and depth of pragmatic reasoning in a controlled laboratory setting, often involving describing one of three objects?

In reference games, what is the speaker's task?

In reference games, what is the listener's task?

Reference games are an instantiation of what type of game from game theory, where a sender knows the state and a receiver does not?

What is the name of the reasoning model that formalizes a hierarchy of pragmatic reasoning types, starting with literal interpreters?

What is the defining behavior of a level-0 sender (S0) in iterated best response models?

What is the defining behavior of a level-0 receiver (R0)?

A level-1 sender (S1), also known as a Gricean speaker, chooses an utterance to maximize what?

A level-2 receiver (R2), also known as a Gricean listener, interprets messages by assuming the speaker is what type?

The interpretation behavior of a level-1 receiver (R1) aligns with what formalization of pragmatic inference popular in theoretical linguistics?

A level-2 sender (S2), or a hyper-pragmatic speaker, chooses an utterance based on the assumption that the listener is what type?

In the 'simple' reference game from Franke & Degen (2016), what choice would a level-0 literal speaker/listener make?

In the 'complex' reference game from Franke & Degen (2016), what choice would a level-1 Gricean speaker (S1) make?

In the 'complex' reference game, what is the only reasoning type that is predicted to reliably choose the target?

What influential probabilistic pragmatics model defines parameterized probabilistic versions of a level-1 Gricean speaker and a level-2 Gricean listener?

In the RSA model, the parameter '\$λ > 0\$' measures, roughly, the speaker's _____.

In the RSA model, as the rationality parameter \$λ\$ approaches infinity (\$λ \to \infty\$), what happens to the speaker's behavior?

How does a Gricean listener (R2) in the RSA model form a posterior belief about the intended referent?

A _____ model assumes that the population of speakers and listeners consists exclusively of Gricean types (S1 and R2).

A _____ model assumes that the population of speakers and listeners is a mix of various reasoning types (e.g., S0, S1, S2 and R0, R1, R2).

Why is Bayesian model comparison important when comparing a simple model to a more complex one?

In the comprehension experiment (Exp. 1), participants chose the target more often in simple implicature trials (77%) than in _____ implicature trials (57%).

In the production experiment (Exp. 2), participants chose the target message more often in simple implicature trials (82%) than in _____ implicature trials (53%).

The experimental results showed significant _____ variability in participants' choice behavior, suggesting the population was not uniform.

At the population-level, the homogeneous Gricean RSA model provided a reasonably good fit to the _____ data.

Why is it problematic to draw conclusions about individuals based on a model that only fits the population's average behavior?

To analyze individual-level data, Franke & Degen (2016) used a _____ modeling approach where each participant was assigned a latent reasoning type.

What does a Bayes factor greater than 1 suggest when comparing a simple model (M_hom) to a complex model (M_het)?

For the production data, the Bayes factor analysis favored a 'charitable' version of the homogeneous model. What does this suggest about speakers?

For the comprehension data, the Bayes factor was approximately \$2.8 \times 10^{-11}\$. What does this overwhelmingly small number indicate?

What was the final, nuanced conclusion of the Franke & Degen (2016) study regarding speaker and listener types?

The comprehension data suggested that the majority of listeners were not Gricean (R2), but rather what type of reasoner?

What level of Theory-of-Mind (ToM) reasoning corresponds to the majority of participants in the study (Gricean S1 speakers and exhaustive R1 listeners)?

What is one key implication of the study's findings for the field of probabilistic pragmatics?

In probabilistic pragmatics, Gricean speakers are typically modeled as level-___ ToM-reasoners.

In probabilistic pragmatics, Gricean listeners are typically modeled as level-___ ToM-reasoners.

What is the term for a speaker who chooses the most informative true utterance assuming the listener is an exhaustive (R1) listener?

The comprehension results showed that participants' performance on complex implicature trials (57% target choice) was significantly above ____.

What potential individual-level cognitive factors might explain differences in pragmatic reasoning types?