Training on the Job: Learning While Searching in an Engineering Workplace Dirk Ahlers Mahsa Mehrpoor NTNU – Norwegian University of Science and Technology Trondheim, Norway dirk.ahlers@idi.ntnu.no, mahsa.mehrpoor@ntnu.no ABSTRACT handle files, we focus on an intranet or internal network file Professional search in an engineering context includes users system as a corpus [1] on top of which we are developing an on different stages from novice to expert. We discuss how expert search system [2]. searching as learning can help to understand searching in a larger information seeking, workspace, and learning environ- 2. WORK, SEARCH, TRAINING ment, giving users the tools and understanding to become In a specialised professional search system that is a large experts over time. We present some findings of user inter- part of daily information seeking activities and work tasks, action with engineering domain retrieval systems where the users gain increasing knowledge and expertise over time by searching as learning perspective can improve our under- interacting with the search system. We argue in this pa- standing of information seeking behaviour. per that the information seeking tasks of engineers working on a complex project can be understood from a searching CCS Concepts as learning perspective. This perspective can help under- •Information systems → Information retrieval; Users and stand the users as it better fits their progress from novice interactive retrieval; •Human-centered computing → to expert, both in the short and long term, it may explain Computer supported cooperative work; certain changes of information seeking behaviour over time, and also inform the adaptability of the system. Previous work has looked at conceptual types of learning as searching Keywords in the workplace [3], here we examine a concrete example. Information Seeking, Information Access, Engineering, Pro- Learning in this search scenario takes place at at least fessional Search two levels. The simpler and more general one is that users learn how to use the system, especially learning how to for- 1. INTRODUCTION mulate queries to better retrieve relevant documents, how In a number of professional search settings the goal is to judge relevance of results, etc. This is basically learn- not only to retrieve a particular isolated fact or document, ing how to search and already a well-explored area. Still, but also to satisfy more complex information needs as part from a research perspective, we are also interested to see of a surrounding work task. In an inherently knowledge- how judgements are made in this context. driven work environment, searching for and working with On a more specific case, users learn where to find doc- documents is a common task that is considered especially uments, which type or source of document is more useful challenging for new personnel in a team or company due to for a certain task, and also learn the relevant documents an often steep learning curve. and processes in their work environment, the unwritten (and In our ongoing work, we examine the use of search, rec- written) rules of the company or the team. Undoubtedly, a ommendation, and knowledge management tools to support learning effect occurs when retrieved documents are studied. engineers [5] in information seeking tasks in their daily work Users internalize gained information and knowledge, chang- [2]. Our users work in larger interdisciplinary teams, de- ing the information needs they will have from the search signing and building complex products, with engineers from system in the future. This last part is the most interesting many different disciplines as well as project managers and from the searching as learning angle. If we take the defini- finance, legal, or HR contributors. Due to the large num- tions of learning as a change in the knowledge structures of ber of files generated and the preferred way of the teams to a user [8], learning should occur for the engineers on their way to higher expertise in their job in general and also in smaller units within a smaller project. The basis for the discussions in this paper are a number of informal interviews and discussions with engineers from a large engineering company that builds large-scale offshore structures [2] and others and more formal interviews with a number of student groups at our university that take part in Search as Learning (SAL), July 21, 2016, Pisa, Italy a challenge to build an energy-efficient race car from scratch. In the former case, team sizes range from around 5 to large The copyright for this paper remains with its authors. Copying permitted for private and academic purposes. international teams of hundreds, in the latter case, there are Information Interaction Process by observing the interaction of experts with a retrieval sys- Work Tasks Seeking Context tem, certain document characteristics and relations as well as interesting workflow aspects could be mined. This proved Information Needs extremely hard, as one of the findings of our interviews was Work process Search Process IR/ that for many tasks, expert searchers use less searching in- Recommender Document System Storage side the system. In short, experts use retrieval systems a lot User Knowledge-Based Engineering System Rules less and with a different focus due to their experience and expertise. This means that they may not need to use search to find the relevant information or that they do know much more facts and processes and thus do not have any informa- tion need at all. Thus simply observing an expert to train Figure 1: Information seeking tasks embedded in or support novices turned out to be not feasible. work tasks with engineering tools integration (Fig- However, another finding is that engineering experts search ure from [2] adapted from [6]) for answers that support more complex or unfamiliar tasks. Because search tasks often consist of multiple linked searches for different snippets of information, this is inherently dif- ficult to model. These findings are consistent with results around 10-15 students per group. We interviewed about 5 from a study of another aspect of engineering support [7]. engineers and about 20 students. They all come from differ- Saved time is not fully assigned to other tasks. Instead, it ent disciplines and have different backgrounds and different is often used for increased breadth and depth of the search roles in the project. for solutions to harder and more complex problems. This Engineering large complex products or projects is inher- basically frees up time and resources to rethink existing so- ently a collaborative exercise. In the student example, learn- lutions or tackle harder issues on higher complexity levels. ing is of course the main desired outcome. Students share knowledge directly, but also search and read documents from their own and previous groups as well as general documen- 3. CONCLUSION tation to advance their knowledge. In the company example We see some angles to explore in future work. It would with a large workplace organization, the same happens, and be interesting to more systematically map previous work [8] novices are also guided by expert engineers. However, an more systematically to deeper explore this context. The additional task there is expert search, in which users search issue of ’good abandonment’ where a result page already for experts in the system to know whom to ask for particular covers the information need and no further interaction takes questions. Information needs also change depending on the place is stronger in this scenario as in many cases, we would work and project context [4]. not even see the user in the system in the first place. A Figure 1 shows a simplified view of an engineer’s work sit- decreased use could also hint towards an advancement of uation, work tasks, resulting information needs, and possible the user who is well on his learning path. Connected to this engagement with a retrieval system from the view of infor- would be a measure of the complexity of the search task mation seeking [6]. Overlayed is the support by certain types to better support users in their different stage of expertise. of engineering tools that may integrate some parts of these This could inform understanding for future retrieval system layers. So-called Knowledge-based engineering (KBE) sys- design in professional search. tems allow construction of parameterised solutions to stan- dard engineering questions [7] and are very knowledge-intensive. 4. REFERENCES Not presented is the social environment consisting of team [1] D. Ahlers and M. Mehrpoor. Everything is filed under members, hierarchical relations, or expert mentors. This re- ‘file’ – conceptual challenges in applying semantic flects concepts of individual, social, and technical dimensions search to network shares for collaborative work. of learning activities [3]. Hypertext ’15. ACM, 2015. One important thing to note is that there is a lot of in- [2] D. Ahlers, M. Mehrpoor, K. Kristensen, and teraction with documents, external Web sources, discussions J. Krogstie. Challenges for information access in with colleagues or project verification and reviews that takes multi-disciplinary product design and engineering place outside the context of the search system and may be settings. In ICDIM. IEEE, 2015. invisible from that system. From a learning perspective, [3] K. Byström. Searching as a learning activity in real life information needs even for the exactly same work task may workplaces. In Searching as Learning Workshop, 2014. change over time. Novices need to find their way around the [4] L. S. Freund. Exploiting task-document relations in system, find out where documents are stored and how they support of information retrieval in the workplace. PhD are organised, and need to familiarise themselves with the thesis, University of Toronto, 2008. relevant standards and processes while experts have learned [5] M. Hertzum and A. M. Pejtersen. The information- a lot of these already. On the other hand, there are certain seeking practices of engineers: searching for documents standards and protocols that need always be followed in the as well as for people. Inf. Process. Manage., 36(5), 2000. latest version for specific parameters or processes. However, this may still mean that users have bookmarked the latest [6] P. Ingwersen and K. Järvelin. The Turn: Integration of source and need not search for it or that they have a physical Information Seeking and Retrieval in Context. 2005. printout in the office for reference. [7] G. L. Rocca. Knowledge based engineering: Between AI Understanding information seeking tasks with the search- and CAD. Adv. Eng. Inform., 26(2):159 – 179, 2012. ing as learning view helps understand some of our initial [8] P. Vakkari. Searching as learning: A systematization findings in a new light. For example, an initial idea that based on literature. J. Inf. Sci., 42(1):7–18, 2016.