How do we use search engines for learning?

SALIENT, an interdisciplinary project, explores how we can use internet search for learning purposes

diesen Beitrag auf Deutsch lesen

We can hardly imagine life without the Web as a search tool, it is simply part of the learning process: whether at school, at work or for recreation. If the Web were a library, it would be the world’s largest – with millions of shelves stacked with countless books, texts, CDs and films.

And yet how can we get to grips with all of this input if we want to learn something? How do we search for information when we want to know something and learn about it? What type of search process leads us to which goal? What action do we take when looking for certain information on a topic, and what significance does information search have when learning? Search engines, link lists, wikis and video portals all offer search assistance. Under the term “search as learning” (SAL), researchers have spent several years investigating the conditions that must be met for these search processes to succeed.

In 2018, the interdisciplinary research project “SALIENT: Search as Learning – Investigating, Enhancing, and Predicting Learning during Multimodal (Web) Search” started exploring how web search can be changed or enhanced to ensure that learning and knowledge acquisition are supported by better results during such search processes. The cooperation partners from the fields of computer science (TIB, L3S Research Center, and GESIS – Institute for the Social Sciences) and psychology (Leibniz-Institut für Wissensmedien – IWM) collaboratively explore how multimedia online resources, such as texts, images and videos, are used to meet learning needs. One example is the use of Web documents that can be retrieved by search engines when tackling learning tasks.

Added benefit for searchers: ranking of more appropriate learning content

The aim of the project is to be able to detect and predict what interests searchers during the actual search process; the results can then be displayed ranked by their significance for the individual user during the search. Another step involves developing “recommender systems”, i.e. systems that recommend more appropriate learning content to users, such as diagrams, lecture slides and videos. As a result, a user’s search query will generate more comprehensive information.

The SALIENT project approaches this issue in several steps: the first step is to determine whether there is actually a learning intention behind a user’s Web search. After all, search engines are used in a variety of contexts – and if the user simply wants to quickly book a flight or a hotel room, they would no doubt find it distracting to get detailed information on the history and significance of the destination. To find out how learners in particular proceed during web search, a large-scale laboratory study was conducted in the context of SALIENT: first of all, over 100 test persons’ knowledge on a clearly defined topic – the development of thunderstorms – was tested. Next, the participants were asked to research the topic in a time-limited web search and to refresh and expand their knowledge. Their knowledge gain was then measured. Since the study was conducted under laboratory conditions, the researchers were able to collect a lot of detailed data about the test persons’ search processes – obviously, the pages visited and the learning resources they contained, and whether they were texts, images or videos, as well as mouse movements and clicks, but also eye-tracking data showing the parts of the websites viewed that test persons looked at in greater detail.

This extensive wealth of data should help answer a wide range of research questions – from a psychology perspective and a computer science perspective. SALIENT has a special focus on multimedia elements and how they are integrated into the learning process. This is achieved by enabling the test subjects to search the web freely. It transpired that they increasingly used multimedia learning materials, especially videos, in their search.

On the basis of short essays written during the knowledge tests, the researchers are able to investigate the participants’ acquisition of terminology – to what extent do they acquire new vocabulary and from which websites did the new terms originate? What influence do images and videos have on learning success compared to textual materials, which were searched more extensively? Can recommendations for the design of effective learning materials be derived from this with regard to their multimodal composition? To what extent do learners’ personal competencies have to be taken into account? And how can search engine rankings be adapted to provide enhanced support?

Although the data has not yet been fully evaluated, initial answers can already be drawn from the findings. For example, the measurement of learning success depends on the way in which knowledge is tested after the learning phase: an examination of the short essays written by the test persons revealed that knowledge gain depended primarily on the texts of the websites they visited, while videos had very little impact. The situation is different when considering the results of multiple-choice tests. In this case, knowledge gain can be better predicted if multimedia features of websites are included. The researchers examined the extent to which a website consisted of text, images and videos, for instance, and the types of images used (photos, diagrams, etc.). They also discovered that searchers prefer different modalities, depending on their prior knowledge. Users with limited prior knowledge, for example, were found to have a preference for videos.

“The promising results with regard to the impact of multimedia elements on learning success now allow us to plan the next steps,” stated Professor Dr. Ralph Ewerth, spokesperson of the SALIENT project and Head of the Visual Analytics Research Group at TIB. “We are currently investigating how the results can be used to improve search engines. We also intend to compare the data we collect with that of search processes for non-learning purposes to get a better understanding of how they differ. Furthermore, we are exploring alternative ranking methods, given that, if the user has a learning intention, the search results must not only be sorted in relation to their search query, but must also take into account which existing materials promise the best learning success,” explained Ewerth.

Tailored learning in the future

The interdisciplinary project makes a major contribution to tomorrow’s innovative library services in the form of virtual learning environments. “Our vision of the future is that, one day, we will be able to provide learners with precise passages from recordings of lectures, as required, such as from TIB’s AV-Portal, or to recommend further teaching material on the web,” says Ewerth.

This article first was published in the TIB Annual Report 2020.