Catherine L. Smith and Soo Young Rieh. 2019. Knowledge-Context in Search Systems: Toward Information-Literate Actions. In Proceedings of the 2019 Conference on Human Information Interaction and Retrieval (CHIIR '19). Association for Computing Machinery, New York, NY, USA, 55–62. https://doi.org/10.1145/3295750.3298940
ABSTRACT
In this perspective paper we define knowledge-context as meta information that searchers use when making sense of information displayed in and accessible from a search engine results page (SERP).
[Mais uma definição sobre Contexto, especÃfica para conhecimento e ferramentas de busca]
We argue that enriching the knowledge-context in SERPs has great potential for facilitating human learning, critical thinking, and creativity by expanding searchers' information-literate actions such as comparing, evaluating, and differentiating between information sources. Thus it supports the development of learning-centric search systems.
[Motivação para incluir contexto em abordagens de busca exploratória. Mas somente contexto da fonte de dados (proveniência)]
Using theories and empirical findings from psychology and the learning sciences, we first discuss general effects of Web search on memory and learning. After reviewing selected research addressing metacognition and self-regulated learning, we discuss design goals for search systems that support metacognitive skills required for long-term learning, creativity, and critical thinking.
[O que já se tem na literatura em outras áreas de conhecimento como psicologia e educação]
We then propose that SERPs make both bibliographic and inferential knowledge-context readily accessible to motivate and facilitate information-literate actions for learning and creative endeavors. A brief discussion of related ideas, designs, and prototypes found in prior work follows. We conclude the paper by presenting future research directions and questions on knowledge-context, information-literate actions, and learning-centric search systems.
[O que propõem para ser acrescentado]
1 Introduction
[Algumas pessoas acreditam que o resultado das buscas são fatos e não questionam as fontes]
Traditionally, Web SERPs have supported these views by conveying meta-information users easily understand in forms such as list order, cue-laden item content summaries (title, url, snippet), entity cards, dictionary definitions, images, video links, news item headlines, items for sale, suggested queries, and extracted facts or answers. Explicit answers are identified directly as such.
Increasingly, Web SERPs also include metadata such as dates of publication, author, publisher, other source attributions, and source-specific metadata such as categories for comic book and movie trivia. All these forms of information comprise the knowledge-context searchers use when making sense of the information displayed on and accessible from the SERP. In a task or answer-centric system, knowledge-context may be designed not only to enable the searcher’s selection of the most relevant and useful results, but also to positively influence the searcher’s confidence in the accuracy and reliability of the system’s comparison, evaluation, and differentiation of the answers, information, and sources presented.
[Metadados de contexto em sistemas de busca como um recurso para aumentar a confiança do usuário na informação, não só para filtrar, agrupar ou correlacionar]
Knowledge-context is an important concept in the field of human information interaction and retrieval because it expands our perspective on the problems we are trying to solve in the design of search and retrieval systems. ... Assuming that task and answer-centric design priorities are important, what might be gained by designing for knowledge-context that induces a user to take information-literate action?
[A busca orientada a tarefa onde a própria tarefa pode trazer restrições de contexto que requeiram filtros e comparações]
2 Knowledge-Context and Information-Literate Actions
In order to learn, understand, and gain confidence in their knowledge, information literate people ask and answer questions about the information they encounter. Information literacy is one of a larger group of literacies that extend from the traditional scope of locating, accessing, and using information to higher order thinking skills and related literacies for technology, visual information, media, and online social engagement [28, 41]. In this paper, we propose that knowledge-context affects ILA associated with learning goals such as the need to remember and synthesize information.
[Alfabetização em informação requer a compreensão do contexto das informações. Usuários com alfabetização em informação fazm perguntas sobre o contexto das informações apresentadas.]
[Definição de ILA: A fim de aprender, entender e ganhar confiança em seu conhecimento, as pessoas competentes em informação perguntam e respondem a perguntas sobre as informações que encontram. O letramento informacional é um de um grupo maior de letramentos que se estende desde o escopo tradicional de localizar, acessar e usar informações até habilidades de pensamento de ordem superior e letramentos relacionados para tecnologia, informação visual, mÃdia e engajamento social online.]
[Não basta encontrar a informação (information findability) é precisa fazer perguntas sobre as informações encontradas para entender o contexto das mesmas]
For example, an information literate person differentiates multiple sources of information on the basis of characteristics such as attribution (who made this claim), the provenance of an information source (e.g., its author, publisher, website owner), the oeuvre of an author, or referents to and from a source (e.g., citations, links, quotations).... Thus ILA is part of the broader concept of sense-making. It is also related to the concept of exploratory search in that it supports learning and investigation. Knowledge-context is readily accessible meta-information, organized and displayed for searchers to use during information seeking.
[As afirmações recuperadas geram perguntas sobre proveniência assim como outras dimensões contextuais. O contexto deve acompanhar a afirmação de modo a responder essas perguntas.]
[ILA é aprendizado auto regulado, habilidade meta cognitiva, conhecimento sobre o próprio conhecimento. Contexto do conhecimento acessÃvel durante as buscas pode incentivar a fazer as perguntas sobre as informações apresentadas.]
3 Effects of Web search on Metacognition
The authors also found evidence that people think of a Web search engine, putatively where they can find information, when asked a trivia question they cannot answer. This result is directly related to the construct feeling-of-knowing, a subjective judgment that one knows something well enough to be able to recall or recognize it in the future [26]. Research on metacognition in learning and psychology often uses feeling-of-knowing as a measure. In research comparing question-answering among people with and without access to a Web search engine, Ferguson, McLean, and Risko [16] found that the feeling-of-knowing an answer is lower when a Web search engine is accessible than when it is not. Lower feeling of knowing is associated with a lower propensity to answer a question from memory.
[Como se a Web fosse uma extensão da memória do indivÃduo, só basta saber procurar. Mas e quando esse conhecimento não está de fato disponÃvel? A sensação é de que não se soube procurar corretamente. Não só a Web mas os aparelhos de telefone com memória fizeram as pessoas não decorarem nem os número de telefone.]
In summary, when people believe information will be stored on a computer they are less likely to remember the information, and more likely to remember where the information has been saved. When people feel they don’t know something, they are mentally primed to search the Web. People make accurate predictions on how easily they can find information on the Web and tend to give up sooner or fail to find what they seek when they foresee difficulty. For topics where information can be found on the Web, the use of Web search leads people to overestimate how much they know, and to generally inflate their judgments on the quality of their own memories and cognitive skills, and on knowing where to find information. These findings have important implications for learning in our Web-centric society.
[Quando as pessoas acreditam que terão dificuldade de encontrar a informação que precisam elas tendem a desistir mais cedo ou realmente não se esforçar para encontrar]
4 Metacognitive Strategies for Self-regulated Learning
One of the fundamental concepts of memory theory is that our memories form and learning occurs when we use what we know by recalling it from our internal memory and actively engaging it in thinking... That is, the acquisition of information with little or no cognitive effort generally results in little or no learning.
[Não existe aprendizado, retenção de informação, sem esforço cognitivo, raciocÃnio e uso de memória]
Metacognitive strategies can help reduce overestimation of knowing and increase the cognitive engagement required for learning. Several strategies with high efficacy involve actively summarizing newly learned information after a delay (long enough to reduce activation in short-term memory)
[Saber não é achar que sabe. Resumir o que aprendeu para reforçar o aprendizado.]
Simply writing five keywords that label what has been learned improves the accuracy for memory of learning (metacomprehension judgements; knowing that you know), as well as for the content of that learning
[ExercÃcio para reter o conhecimento e teste de aprendizado para avaliar o sucesso da tarefa de exploração]
Another highly efficacious metacognitive strategy involves spacing and interleaving during learning, where each iteration between related focal topics requires a shift in attention and recall of the information to be learned. This strategy slows learning in the short term but enhances long-term retention, possibly drawing attention to associations among topics for greater synthesis.
[Refinar uma consulta gera aprendizado pq exige um esforço de reajustar o enfoque da consulta. A eficiência de achar A RESPOSTA rapidamente pode atrapalhar a eficácia para o processo de aprendizado.]
5 Learning-Centric Knowledge-Context
The findings discussed above motivate consideration of the goals and priorities of alternative designs for Web search systems
A learning-centric search system supports knowledge-context as a central design priority. In such a system, searchers encounter an enriched knowledge-context that prompts and facilitates metacognitive engagement, including the metacognitive strategies of information literacy and information literate action. This in turn supports active engagement with information and resulting long-term learning and the metacomprehension needed for successful self-regulated learning.
[Metacognição ... metacognição é aprender sobre como funciona o processo de aprendizagem. Habilidade: Visual, Cinestésica (escrever), ... Estratégia: Como aprender]
Information literate searchers tend to ask a series of questions, such as “Who said that?”; “When did they say it?”; “How have others used it?”; and “How does it compare with statements elsewhere?” The results above suggest that those with access to highly reliable transactive Web memory are likely to trust the Web search system as a source of valid information, believing it obviates any need to ask such questions.
[Perguntas sobre o contexto das afirmações. Aprender e reter o conhecimento com as perguntas.]
In a system designed to impart high levels of confidence, a SERP with minimal knowledge-context provides few if any cues to possible problems with routine, undifferentiated trust in the system and its sources. Hence it offers little that might prompt ILA. Using such a system, information literate searchers may use iterative exploratory search to take actions such as comparing multiple information sources, critically evaluating information, and differentiating to extract information from sources.
[Como contornar a falta de contexto no resultado das máquinas de busca? Fazendo novas consultas, mudando as perguntas]
Readily accessible and navigable knowledge-context facilitates these actions. For instance, a searcher may pay attention to an author or publisher name to make judgments about a source’s authority or credibility. A scholar may judge information credibility by the number and quality of citations to a scholarly article. A student may check their understanding by reviewing documents that present the same information at different reading levels. A journalist may read multiple documents written in the same era in order to examine contrasting viewpoints about a topic. In all of these cases, the searcher seeks to make sense of information by using additional contextual information associated with a source and its content. Assessing credibility, checking understanding, and comparing multiple viewpoints are examples of ILA that knowledge-context can promote and facilitate.
[Exemplo do Sérgio dos periódicos, eventos e autores considerados de referência na área de BD. O modelo de confinaça para avaliar a veracidade da informação é do usuário e pode ser influenciado pela tarefa.]
Information literacy involves the capacity to evaluate critically the system itself and take control of learning by asking and answering questions actively. Information literacy involves not just active engagement with information but also a comprehensive understanding of it. From the viewpoint of information literacy, SERPs enriched with knowledge-context offer rich cues about information from multiple dimensions, allowing searchers to engage deeply with information. In our view, knowledge-context serves as more than raw material and metadata “out there to be accessed efficiently”.
[Contexto como dimensões para avaliar as informações]
Critical thinking is a self-directed thought process that improves an individual’s ability to uncover the deep structure of knowledge by recognizing the relationships between concepts. In fact, the ability to identify and articulate inferential connections between multiple pieces of information is a core concept of critical thinking. Although there are numerous definitions of critical thinking, most concur that it is a higher-order cognitive process that leads to actions such as discovering, sorting, distinguishing, contrasting, integrating, aggregating, synthesizing, and generating.
[Em um KG o relacionamento entre os conceitos está explÃcitamente representado e novos relacionamentos podem ser inferidos. Diferentes ações a serem feitas com a informação para que seja possivel extrair mais informação. Em relação as afirmações, o contexto pode evidenciar relacionamentos que não são visÃveis em KGs tradicionais como por exemplo o compartilhamento de contexto temporal ou espacial entre eventos]
We believe that readily available knowledge-context in SERPs would motivate and facilitate development of critical thinking as searchers engage with information more actively and critically.
Knowledge-context is also important to the creativity process Sawyer describes. In his definition, creative thinking involves recognizing a problem or anomaly, active learning of new information, time for new associations to strengthen in memory, and the generation of ideas. Learning and idea incubation result in a final external expression of a solution or decision. Easy, reliable access to enriched knowledge-context, along with appropriate cues, are likely to provide opportunities for searchers to comprehend multiple aspects of information.
[Aprendizado por associação]
For purposes of our argument, we divide knowledge-context into two types: bibliographic knowledge-context and inferential knowledge-context. The first type focuses on sources, with the goal ... to determine information credibility. ... The second type of knowledge-context pertains to the characteristics of information content, and how content connects one source to another, or affects past or current meanings of a source.
[Existem outras dimensões contextuais como espacial, temporal, temática, ... além da proveniência quando tratamos afirmações e não só documentos]
6 Components of a Learning-Centric Search System
Vakkari and Huuskonen’s examination of student essays revealed that the precision of search results correlated negatively with long-term learning. Students who tended to use the system’s advanced search features in more frequent search sessions experienced lower retrieval precision but learned more than students who used more precise general search features in fewer sessions.
[Aprender é memorizar ou é raciocinar usando as informações que temos? ]
Dörk envisions slow technology as an opportunity to design new interfaces and visualizations that facilitate the use of knowledge-context in activities such as orienteering, exploration, browsing, and the collection of information for later use. Specific design goals in this work focus on intuitive visualizations that integrate explicit and implicit relations in a navigable structure.
[Interface com recursos como os grafos e interações como drill down e drill up interferem no aprendizado?]
Knowledge organization for knowledge-context.
Two recent papers express other ideas that parallel our conception of a learning-centric search system and knowledge-context. In a straightforward example of knowledge-context, Fuhr et al. [18] proposed a set of computable evaluative measures and criteria for metadata describing characteristics of textual news sources: factuality, readability, virality, emotion (valence), opinion, controversy, authority/credibility/trust, technicality, and topicality. They proposed the measures as descriptions of computational characteristics and judgments for a source. The authors present a schema in the style of a nutrition label that would be accessible alongside a source. Accessing and reading the label would be an information-literate action.
[Metáfora de um rótulo nutricional que seria acessÃvel ao lado de uma fonte textual de informações. Acessar e ler o rótulo seria uma ação de alfabetização informacional]
In an inferential form of knowledge-context, Voskarides et al. [50] generated rankings for entities associated on a knowledge graph, with the goal of organizing relevant factual associations and entities for navigation of a large, complex graph. As knowledge-context, the rankings are useful for presenting entity cards relevant to an initial query or search session. Once ranked, the cards may be presented or organized for browsing. We envision a system designed to facilitate interaction with contextualized entity cards and the creation of personal entity cards that restate newly learned information in a personal knowledge-context, discussed next.
[Trabalho com KG, onde o contexto seriam cartões para organizar a navegação no grafo]
Recent work on a learning-centric search system prototype, SearchAssist [23], draws attention to the need for two sources for knowledge-context: a public knowledge-context derived from external information sources accessed via the Web, and a private knowledge-context derived from internal information sources comprising a user’s personal information management (PIM) space...
[Contexto das fontes e contexto do usuário]
In summary, knowledge-context is more than metadata or a knowledge-graph. It is also an organizing schema that makes knowledge structures useful for learning.
[Contexto em um KG permite que o usuário avalie a veracidade e a utilidade das informações para a tarefa em questão]
Interface design.
These include preview, primary view, overview, and review, with look ahead, peripheral, and shared views. The design objective was to guide visual attention during partition, exploration, and sensemaking for a large corpus. .... “One challenge ... is how to show context without complicating the screen display.” For us, the compelling idea in Agileviews is the prospect of gaining a traversable overview that enables visualization of selected associations that form usable partitions within and across forms of knowledge-context.
[Contexto pode sobrecarregar a visualização das informações]
7 Conclusion and Future Research Directions
This paper makes a contribution to the field of human information interaction and retrieval, as well as searching as learning, by proposing a new perspective in which we emphasize the importance of enriching the knowledge-context of retrieval results
[O contexto enriquece os resultados para facilitar o aprendizado]
To make more than a series of marginal improvements to existing systems, collaborative, interdisciplinary research programs on retrieval and interaction design as well as controlled experimental research on the above questions and hypotheses are essential. Such studies require experimental systems and both laboratory and longitudinal research designs. The object of this work is not improved system performance but a system that reflects psychological factors that affect and are affected by knowledge organization, interface design, and the role of speed in interaction.
[O foco não é a eficiência da recuperação e sim a eficácia no processo de busca exploratória para realizar uma tarefa]
[18] N. Fuhr, A. Giachanou, G. Grefenstette, I. Gurevych, A. Hanselowski, K. Jarvelin, R. Jones, et al.. 2018. An Information Nutritional Label for Online Documents. ACM SIGIR Forum, 51(3), 46–66.
[50] N. Voskarides, E. Meij, R. Reinanda, A. Khaitan, M. Osborne, G. Stefanoni, P. Kambadur, and M. de Rijke. 2018. Weakly-supervised Contextualization of Knowledge Graph Facts. In The 41st International ACM SIGIR Conference on Research & Development in Information Retrieval (SIGIR '18). ACM, New York, NY, USA, 765-774.
A ref [50] foi lida, comentários aqui -> https://versant-pesquisadedoutorado.blogspot.com/2022/08/weakly-supervised-contextualization-of.html
ResponderExcluir