Ideology, communication and polarization Kashima, Yoshihisa; Perfors, Andrew; Ferdinand, Vanessa ...
Philosophical transactions of the Royal Society of London. Series B. Biological sciences,
04/2021, Letnik:
376, Številka:
1822
Journal Article
Recenzirano
Odprti dostop
Ideologically committed minds form the basis of political polarization, but ideologically guided communication can further entrench and exacerbate polarization depending on the structures of ...ideologies and social network dynamics on which cognition and communication operate. Combining a well-established connectionist model of cognition and a well-validated computational model of social influence dynamics on social networks, we develop a new model of ideological cognition and communication on dynamic social networks and explore its implications for ideological political discourse. In particular, we explicitly model ideologically filtered interpretation of social information, ideological commitment to initial opinion, and communication on dynamically evolving social networks, and examine how these factors combine to generate ideologically divergent and polarized political discourse. The results show that ideological interpretation and commitment tend towards polarized discourse. Nonetheless, communication and social network dynamics accelerate and amplify polarization. Furthermore, when agents sever social ties with those that disagree with them (i.e. structure their social networks by homophily), even non-ideological agents may form an echo chamber and form a cluster of opinions that resemble an ideological group. This article is part of the theme issue 'The political brain: neurocognitive and computational mechanisms'.
The impressive recent performance of large language models has led many to wonder to what extent they can serve as models of general intelligence or are similar to human cognition. We address this ...issue by applying GPT-3.5 and GPT-4 to a classic problem in human inductive reasoning known as property induction. Over two experiments, we elicit human judgments on a range of property induction tasks spanning multiple domains. Although GPT-3.5 struggles to capture many aspects of human behavior, GPT-4 is much more successful: for the most part, its performance qualitatively matches that of humans, and the only notable exception is its failure to capture the phenomenon of premise non-monotonicity. Our work demonstrates that property induction allows for interesting comparisons between human and machine intelligence and provides two large datasets that can serve as benchmarks for future work in this vein.
•Apply GPT-3.5 and GPT-4 to a classic inductive reasoning task known as property induction.•GPT-3.5 struggles to capture many aspects of human behavior.•GPT-4’s performance qualitatively matches that of humans, aside from the phenomenon of premise non-monotonicity.•Provide two large property induction datasets that can serve as future benchmarks.
A core assumption of many theories of development is that children can learn indirectly from other people. However, indirect experience (or testimony) is not constrained to provide veridical ...information. As a result, if children are to capitalize on this source of knowledge, they must be able to infer who is trustworthy and who is not. How might a learner make such inferences while at the same time learning about the world? What biases, if any, might children bring to this problem? We address these questions with a computational model of epistemic trust in which learners reason about the helpfulness and knowledgeability of an informant. We show that the model captures the competencies shown by young children in four areas: (1) using informants’ accuracy to infer how much to trust them; (2) using informants’ recent accuracy to overcome effects of familiarity; (3) inferring trust based on consensus among informants; and (4) using information about mal‐intent to decide not to trust. The model also explains developmental changes in performance between 3 and 4 years of age as a result of changing default assumptions about the helpfulness of other people.
A core assumption of many theories of development is that children can learn indirectly from other people. However, indirect experience (or testimony) is not constrained to provide veridical information. As a result, if children are to capitalize on this source of knowledge, they must be able to infer who is trustworthy and who is not. How might a learner make such inferences while at the same time learning about the world? What biases, if any, might children bring to this problem?
One of the main limitations of natural language‐based approaches to meaning is that they do not incorporate multimodal representations the way humans do. In this study, we evaluate how well different ...kinds of models account for people's representations of both concrete and concepts. The models we compare include unimodal distributional linguistic models as well as multimodal models which combine linguistic with perceptual or affective information. There are two types of linguistic models: those based on text corpora and those derived from word association data. We present two new studies and a reanalysis of a series of previous studies. The studies demonstrate that both visual and affective multimodal models better capture behavior that reflects human representations than unimodal linguistic models. The size of the multimodal advantage depends on the nature of semantic representations involved, and it is especially pronounced for basic‐level concepts that belong to the same superordinate category. Additional visual and affective features improve the accuracy of linguistic models based on text corpora more than those based on word associations; this suggests systematic qualitative differences between what information is encoded in natural language versus what information is reflected in word associations. Altogether, our work presents new evidence that multimodal information is important for capturing both and concrete words and that fully representing word meaning requires more than purely linguistic information. Implications for both embodied and distributional views of semantic representation are discussed.
Most studies of ambiguity aversion rely on experimental paradigms involving monetary bets. Thus, the extent to which ambiguity aversion occurs outside of such contexts is much less understood, ...particularly when the situation cannot easily be reduced to numerical terms. The present work seeks to understand whether people prefer to avoid ambiguous decisions in a variety of different qualitative domains (e.g., work, family, love, friendship, exercise, study, and health), and, if so, to determine the role played by prior beliefs in those domains. Across three studies, we presented participants with 24 vignettes and measured the degree to which they preferred risk to ambiguity in each. We also asked them for their prior probability estimates about the likely outcomes in the ambiguous events. Ambiguity aversion was observed in the vast majority of vignettes, but at different magnitudes. It was predicted by whether the vignette involved gain or loss as well as by people's prior beliefs; however, the heterogeneity between people meant that the role of prior beliefs was only evident in an individual‐level analysis (i.e., not at the group level). Our results suggest that the desire to avoid ambiguity occurs in a wide variety of qualitative contexts but to different degrees for different people and may be partially driven by unfavorable prior estimates of the likely outcomes of the ambiguous events.
Human languages vary in many ways but also show striking cross‐linguistic universals. Why do these universals exist? Recent theoretical results demonstrate that Bayesian learners transmitting ...language to each other through iterated learning will converge on a distribution of languages that depends only on their prior biases about language and the quantity of data transmitted at each point; the structure of the world being communicated about plays no role (Griffiths & Kalish, , ). We revisit these findings and show that when certain assumptions about the relationship between language and the world are abandoned, learners will converge to languages that depend on the structure of the world as well as their prior biases. These theoretical results are supported with a series of experiments showing that when human learners acquire language through iterated learning, the ultimate structure of those languages is shaped by the structure of the meanings to be communicated.
We introduce a new resource: the SAYCam corpus. Infants aged 6–32 months wore a head-mounted camera for approximately 2 hr per week, over the course of approximately two-and-a-half years. The result ...is a large, naturalistic, longitudinal dataset of infant- and child-perspective videos. Over 200,000 words of naturalistic speech have already been transcribed. Similarly, the dataset is searchable using a number of criteria (e.g., age of participant, location, setting, objects present). The resulting dataset will be of broad use to psychologists, linguists, and computer scientists.
When people use samples of evidence to make inferences, they consider both the sample contents and how the sample was generated (“sampling assumptions”). The current studies examined whether people ...can update their sampling assumptions – whether they can revise a belief about sample generation that is discovered to be incorrect, and reinterpret old data in light of the new belief. We used a property induction task where learners saw a sample of instances that shared a novel property and then inferred whether it generalized to other items. Assumptions about how the sample was selected were manipulated between conditions: in the property sampling frame condition, items were selected because they shared a property, while in the category sampling frame condition, items were selected because they belonged to a particular category. Experiment 1 found that these frames affected patterns of property generalization regardless of whether they were presented before or after the sample data was observed: in both cases, generalization was narrower under a property than a category frame. In Experiments 2 and 3, an initial category or property frame was presented before the sample, and was later retracted and replaced with the complementary frame. Learners were able to update their beliefs about sample generation, basing their property generalization on the more recent correct frame. These results show that learners can revise incorrect beliefs about data selection and adjust their inductive inferences accordingly.
•Constraints on sampling (“sampling frames”) affect inductive generalization.•Sampling frames affect induction when they are encountered before or after the data.•People can flexibly shift to new sampling frames when old frames are retracted.
Inductive learning is impossible without overhypotheses, or constraints on the hypotheses considered by the learner. Some of these overhypotheses must be innate, but we suggest that hierarchical ...Bayesian models can help to explain how the rest are acquired. To illustrate this claim, we develop models that acquire two kinds of overhypotheses – overhypotheses about feature variability (e.g. the shape bias in word learning) and overhypotheses about the grouping of categories into ontological kinds like objects and substances.
In describing how people generalize from observed samples of data to novel cases, theories of inductive inference have emphasized the learner's reliance on the contents of the sample. More recently, ...a growing body of literature suggests that different assumptions about how a data sample was generated can lead the learner to draw qualitatively distinct inferences on the basis of the same observations. Yet, relatively little is known about how and when these two sources of evidence are combined. Do sampling assumptions affect how the sample contents are encoded, or is any influence exerted only at the point of retrieval when a decision is to be made? We report two experiments aimed at exploring this issue. By systematically varying both the sampling cover story and whether it is given before or after the training stimuli we are able to determine whether encoding or retrieval issues drive the impact of sampling assumptions. We find that the sampling cover story affects generalization when it is presented before the training stimuli, but not after, which suggests that sampling assumptions are integrated during encoding.