• 2019-2023close
• Publications close
• Article close
• Digitala Vetenskapliga Arkivet - Ac... close
• Publikationer från KTH close
• Digital Humanities and Cultural Her... close

Educational content recommendation is a cornerstone of AI-enhanced learning. In particular, to facilitate navigating the diverse learning resources available on learning platforms, methods are needed for automatically linking learning materials, e.g., in order to recommend textbook content based on exercises. Such methods are typically based on semantic textual similarity (STS) and the use of embeddings for text representation. However, it remains unclear what types of embeddings should be used for this task. In this study, we carry out an extensive empirical evaluation of embeddings derived from three different types of models: (i) static embeddings trained using a concept-based knowledge graph, (ii) contextual embeddings from a pre-trained language model, and (iii) contextual embeddings from a large language model (LLM). In addition to evaluating the models individually, various ensembles are explored based on different strategies for combining two models in an early vs. late fusion fashion. The evaluation is carried out using digital textbooks in Swedish for three different subjects and two types of exercises. The results show that using contextual embeddings from an LLM leads to superior performance compared to the other models, and that there is no significant improvement when combining these with static embeddings trained using a knowledge graph. When using embeddings derived from a smaller language model, however, it helps to combine them with knowledge graph embeddings. The performance of the best-performing model is high for both types of exercises, resulting in a mean Recall@3 of 0.96 and 0.95 and a mean MRR of 0.87 and 0.86 for quizzes and study questions, respectively, demonstrating the feasibility of using STS based on text embeddings for educational content recommendation. The ability to link digital learning materials in an unsupervised manner—relying only on readily available pre-trained models—facilitates the development of AI-enhanced learning.

Eelco Runia’s view of historical time is that the now and the past exert an alternating influence on each other. The most prominent expression of this is a dialogue between the historian and the historical actors in which the historian creates metonyms, replacement terms, for how historical actors are defined, which affects the description of the past. In this paper, Runia’s conception of historical time has been tested through an empirical investigation into the changing descriptions of the process of rejecting the aether hypothesis, 1858-2022. After the process of phasing the aether out of physics had previously been explained as being caused by the gradual acceptance of the theory of relativity, the cause was suddenly superseded in the 1990s by the Michelson-Morley experiment conducted in the 1880s. During the 21st century, however, the aether once again returned as a concept. The reason for the changes, is the conclusion, could possibly be due to the ambition to give a rational picture of the development of natural science at every point in time

This paper presents results from a large-scale study of history teachers in Swedish secondary schools. The study examines perceptions of history, content being taught, teaching methods and use of digital technology. The study uses the Technological Pedagogical and Content Knowledge (TPACK) framework to analyse the results together with narrative theory. The main results indicate that knowledge of the past and contemporary perspectives from a canonical tradition are prioritised, together with a content-based lecture-style pedagogy. The use of digital technology does not seem to challenge methods or bring new perspectives to history teaching. However, to fully understand history teaching with technology, a framework that emphasises all parts of history education is needed. This calls for further development of the TPACK model, which is further discussed in this article.

The main target of this study is to make a comparative analysis between the presence or absence of historical thinking competencies in the official curricula of Gymnasieskola (Sweden) and Bachillerato (Spain). To reach this goal, a mixed instrument was designed, both qualitative and quantitative, to analyze the offical documents. The results show a great presence of “historical consciousness” in the case of Sweden and the use of historical sources and evidences in the case of Spain.

Abstract Introduction Public perception of the seriousness of the COVID-19 pandemic compared to six other major public health problems (alcoholism and drug use, HIV/AIDS, malaria, tuberculosis, lung cancer and respiratory diseases caused by air pollution and smoking, and water-borne diseases like diarrhea) is unclear. We designed a survey to examine this issue using YouGov’s internet panels in seven middle-income countries in Africa, Asia, and Latin America in early 2022. Methods Respondents rank ordered the seriousness of the seven health problems using a repeated best-worst question format. Rank-ordered logit models allow comparisons within and across countries and assessment of covariates. Results In six of the seven countries, respondents perceived other respiratory illnesses to be a more serious problem than COVID-19. Only in Vietnam was COVID-19 ranked above other respiratory illnesses. Alcoholism and drug use was ranked the second most serious problem in the African countries. HIV/AIDS ranked relatively high in all countries. Covariates, particularly a COVID-19 knowledge scale, explained differences within countries; statistics about the pandemic were highly correlated with differences in COVID-19’s perceived seriousness. Conclusions People in the seven middle-income countries perceived COVID-19 to be serious (on par with HIV/AIDS) but not as serious as other respiratory illnesses. In the African countries, respondents perceived alcoholism and drug use as more serious than COVID-19. Our survey-based approach can be used to quickly understand how the threat of a newly emergent disease, like COVID-19, fits into the larger context of public perceptions of the seriousness of health problems.

This article takes a comparative approach to settler colonial violence in the American Southwest and German Southwest Africa. The Anglo invasion of central Arizona in 1864 and the German conflict against the Herero in 1904 highlights the nature of frontier violence and identifies similarities and differences across two points in space and time that have seldom been compared by historians. Those writing of the US-Apache conflicts have failed to look to colonial theaters around the world, their transnational attention focusing instead on the borderlands of United States, Mexico, and independent Indians. Similarly, research on the violence in GSWA has not engaged systematically with international parallels and has instead focused on identifying possible links between GSWA and the Nazis and the Holocaust. This article seeks to address these shortcomings by analysing the comparative strands of settler colonial violence.

Special issue

The following study describes a data-driven learning scenario aimed at raising sociolinguistic awareness of matters related to gender, language and stereotyping. The design is inspired by the matched-guise technique (MGT), a quantitative data driven experimental method that has been used extensively to investigate language attitudes. In the scenario, differences in respondents’ response patterns to two gender-manipulated versions (male-female vs. female-male dyads) of the same recorded dialogue were used as a starting point for awareness-raising activities aimed at highlighting how gender stereotypes may affect perceptions of a dialogue. The main focus of the article is a comparison of the learning outcomes of two variants of the setup: a traditional undisclosed MGT-inspired setup, where the design and purpose of the experiment was kept secret until after the response phase, and a so-called open-guise design, where respondents were informed of the design and purpose of the experiment prior to the response phase. Preliminary results suggest that respondents adjust their assessments of a speaker depending on the guise, even when they know it is the same speaker they are listening to. Moreover, the open-guise design seemed to lead to greater pedagogic impact than the scenario based on the undisclosed design. However, further studies are needed to confirm these findings.