More Like this

1. Two kinds of polar knowledge

   https://doi.org/10.1080/10899995.2020.1838849  

   Hamilton, Lawrence C. (November 2020, Journal of Geoscience Education)

   null (Ed.)

   Outreach and communication with the public have substantial value in polar research, in which studies often find changes of global importance that are happening far out of sight from the majority of people living at lower latitudes. Seeking evidence on the effectiveness of outreach programs, the U.S. National Science Foundation sponsored large-scale survey assessments before and after the International Polar Year in 2007/2008. Polar-knowledge questions have subsequently been tested and refined through other nationwide and regional surveys. More than a decade of such work has established that basic but fairly specific knowledge questions, with all answer choices sounding plausible but one being uniquely correct, can yield highly replicable results. Those results, however, paint a mixed picture of knowledge. Some factual questions seem to be interpreted by many respondents as if they had been asked for their personal beliefs about climate change, so their responses reflect sociopolitical identity rather than physical-world knowledge. Other factual questions, by design, do not link in obvious ways to climate-change beliefs—so responses have simpler interpretations in terms of knowledge gaps, and education needs. 

   more » « less
2. Following science on social media: The effects of humor and source likability

   https://doi.org/10.1177/0963662520986942  

   Yeo, Sara K.; Cacciatore, Michael A.; Su, Leona Yi-Fan; McKasy, Meaghan; O’Neill, Liane (January 2021, Public Understanding of Science)

   null (Ed.)

   Science communicators have been encouraged to use humor in their online engagement efforts. Yet, humor’s effectiveness for engaging people with science remains an open question. We report the results of an experiment designed to elicit varied levels of mirth in respondents, which was positively associated with perceived likability of the communicator and motivation to follow more science on social media. Furthermore, mirth and perceived likability serially mediated the effect of the experimental manipulation on motivation and factual science knowledge served as a moderator. This indicates that, while humor might be an effective means of reaching audiences, downstream effects are likely to vary depending on individuals’ knowledge. 

   more » « less
3. Diverse and faithful knowledge-grounded dialogue generation via sequential posterior inference

   Xu, Y.; Kong, D.; Xu, D.; Ji, Z.; Pang, B.; Fung, P.; Wu, Y. N. (October 2023, International Conference on Machine Learning (ICML 2023))

   The capability to generate responses with diversity and faithfulness using factual knowledge is paramount for creating a human-like, trustworthy dialogue system. Common strategies either adopt a two-step paradigm, which optimizes knowledge selection and response generation separately and may overlook the inherent correlation between these two tasks, or leverage conditional variational method to jointly optimize knowledge selection and response generation by employing an inference network. In this paper, we present an end-to-end learning framework, termed Sequential Posterior Inference (SPI), capable of se- lecting knowledge and generating dialogues by approximately sampling from the posterior distribution. Unlike other methods, SPI does not require the inference network or assume a simple geometry of the posterior distribution. This straightforward and intuitive inference procedure of SPI directly queries the response generation model, allowing for accurate knowledge selection and generation of faithful responses. In addition to modeling contributions, our experimental results on two common dialogue datasets (Wizard of Wikipedia and Holl-E) demonstrate that SPI outperforms previous strong baselines according to both automatic and human evaluation metrics. 

   more » « less
4. Knowledge Card: Filling LLMs' Knowledge Gaps with Plug-in Specialized Language Models

   Feng, Shangbin; Shi, Weijia; Bai, Yuyang; Balachandran, Vidhisha; He, Tianxing; Tsvetkov, Yulia (May 2024, International Conference on Learning Representations)

   By design, large language models (LLMs) are static general-purpose models, expensive to retrain or update frequently. As they are increasingly adopted for knowledge-intensive tasks, it becomes evident that these design choices lead to failures to generate factual, relevant, and up-to-date knowledge. To this end, we propose Knowledge Card, a modular framework to plug in new factual and relevant knowledge into general-purpose LLMs. We first introduce knowledge cards---specialized language models trained on corpora from specific domains and sources. Knowledge cards serve as parametric repositories that are selected at inference time to generate background knowledge for the base LLM. We then propose three content selectors to dynamically select and retain information in documents generated by knowledge cards, specifically controlling for relevance, brevity, and factuality of outputs. Finally, we propose two complementary integration approaches to augment the base LLM with the (relevant, factual) knowledge curated from the specialized LMs. Through extensive experiments, we demonstrate that Knowledge Card achieves state-of-the-art performance on six benchmark datasets. Ultimately, Knowledge Card framework enables dynamic synthesis and updates of knowledge from diverse domains. Its modularity will ensure that relevant knowledge can be continuously updated through the collective efforts of the research community. 

   more » « less
5. Developing and using a scalable assessment to measure preservice elementary teachers' content knowledge for teaching about matter

   https://doi.org/10.1002/tea.21894  

   Castellano, Katherine_E; Mikeska, Jamie_N (September 2023, Journal of Research in Science Teaching)

   Abstract There is strong agreement in science teacher education of the importance of teachers' content knowledge for teaching (CKT), which includes their subject matter knowledge and their pedagogical content knowledge. However, there are limited instruments that can be easily administered and scored on a large scale to assess and study elementary science teachers' CKT. Such measures would support strategic monitoring of large groups of science teachers' CKT and the investigation of comparative questions about science teachers' CKT longitudinally across the professional continuum or across teacher education or professional development sites. To address this gap, this study focused on designing an automatically scorable summative assessment that can be used to measure preservice elementary teachers' (PSETs') CKT in one high‐leverage science content area: matter and its interactions. We conducted a field test of this CKT instrument with 822 PSETs from across the United States and used the response data to examine how this instrument functions as a potential tool for measuring PSETs' CKT in this science content area. Results suggest this instrument is reliable and can be used on large scale to support valid inferences about PSETs' CKT in this content area. In addition, the dimensionality analysis showed that all items measure a single construct of CKT about matter and its interactions, as participants did not show any differential performance by content topic or work of teaching science instructional tool categories. Implications for progressing the field's understanding of the nature of CKT and approaches to developing summative instruments to assess science teachers' CKT are discussed. 

   more » « less