 Research
|
Morten Hertzum |
|
My research interests are within human-computer interaction (HCI), computer supported cooperative work (CSCW), information seeking, and health informatics. More specifically, my research has concerned the following topics.
Usability and user experience
Usability. The term usability is ubiquitous in human-computer interaction. Rather than one established meaning of usability, there are, however, multiple images of usability, each providing a partial view. We have described six images of usability: universal usability, situational usability, perceived usability, hedonic usability, organizational usability, and cultural usability (IJHCI2010). Collectively, these images make usability a sensitizing concept (HCI2018). For the three aspects of usability in the ISO9241-11 definition we have found that information about one aspect - such as efficiency - does not enable us to say much about other aspects - such as effectiveness and satisfaction (CHI2000). Based on repertory-grid interviews we have investigated the personal constructs people employ in talking about their experiences using systems; that is, their personal usability constructs. We have done this for users and deveopers (IJHCI2011), for usability professionals (IJHCS2012), and for all three groups combined (INTERACT2013).
User experience. Two vocal accounts of the factors that shape the adoption and use of information technology are the technology acceptance model (TAM) and work on user experience (UX) models. To explore the experiential component in human-computer interactions we have reviewed papers in the overlap between TAM and UX models (TOCHI2017). Some products, such as robotic vacuum cleaners, transform an activity even though they provide inferior user experiences (IWC2024). We have also studied UX in a work setting and found that ordinary user experiences dominate at work (TOCHI2020). A distinct component of user experience is workload, which is often measured with the six subscales of the task load index (TLX). We have provided reference values and subscales patterns for this self-report instrument of workload (Ergonomics2021). We have also reviewed the associations between TLX and performance (BIT2022) and shown that users more consistently and more strongly prefer systems that impose lower workload than better performance (IJHCS2025).
Technostress. In a follow-up on a seminal study from 2004, we have found that frustration is still a common user experience (TOCHI2023). The participants in our study reported losing 11-20% of the one-hour reporting period to frustrating incidents with their computer. Such frustrating incidents cause technostress. We find that as much as 66% of the incidents experienced by the staff at a hospital department could merely be solved for now, that is, the concrete situation is resolved but the underlying issue remains unsolved (IJMI2024). As a consequence, most of the stressful situations are likely to recur later.
Usability testing. Usability testing is easy to do, but difficult to do well; we have published a practitioner's guide (M&C2020). We recommend heeding the focus of usability tests on concrete use and, therefore, discourage thinking about usability tests as a kind of interview (Interactions2016). Recently, crowdsourcing has made unmoderated usability sessions possible, and they appear to be similar to moderated sessions in terms of the content of the users' verbalizations (IJHCI2015) and the number of usability issues encountered (BIT2014). The users can verbalize their thoughts while using the tested system or after using it; we have reviewed what is gained and lost by using one or the other of these two possibilities (TOCHI2024). In addition to the users' verbalizations, the moderator talks quite a bit during test sessions that employ relaxed thinking aloud (NordiCHI2018). If the focus is on organizational usability then usability tests must leave the lab (JUS2018).
Threats to the quality of usability evaluations. Usability evaluation methods are central to user-centered design and must themselves be thoroughly evaluated. We have found that evaluators who evaluate the same system using the same usability evaluation method tend to detect substantially different sets of problems in the system. This evaluator effect has been documented for usability tests, cognitive walkthroughs, and heuristic evaluations (CHI1998, HFES1998, HCII1999, IJHCI2003, CHI2002, BIT2014). The severity of usability problems is also difficult to assess reliably, but important to the prioritization of problems and the impact of evaluations (IJHCI2006). Furthermore, we see indications that relaxed thinking aloud, which is widely used, affects behaviour and mental workload (BIT2009), and that classic thinking aloud is reactive in the presence of interruptions (IJHCI2013b) and influences perceived time (HF2015). As a supplementary measure of mental workload we have investigated the ratio of perceived time to clock time (IJHCI2013a).
Cultural usability. The production, evaluation, and use of information technologies are international activities, but there are indications that prominent usability evaluation methods developed in Europe and the
Ease of use vis-à-vis security. One particular tension in the concept of usability is between ease of use and security. This tension has led to an interest in usable security, which is about matching security principles and demands against user knowledge and motivation. At the concrete level, we have assessed the security and ease of use of Danish e-banking systems (AJIS2004) and worked with minimal-feedback hints intended to help people remember their passwords (Interactions2006).
Information seeking
Collaborative information seeking. There is compelling empirical evidence of the collaborative nature of much information seeking, but most retrieval systems and many information-science models still depict information seeking as an activity performed by individuals. Collaborative information seeking can be defined as the combined activity of information seeking and collaborative grounding (IPM2008). A field study in the film-archive domain illustrated that collaborative information seeking is common and discussed the prospects of supporting it by means of a collaboratory (CoLIS2002). A study of the medication process showed that most breakdowns in collaborative information seeking in this domain are breakdowns in collaborative grounding (IPM2010). We have also investigated the links between collaborative information seking and organizational procedures (CIS2015) and between collaborative information seeking and expertise seeking (JDOC2017). The empirical literature on collaborative information seeking provides data about behavior to a larger extent than reasons, experiences, and especially outcomes (JDOC2019).
Expertise seeking. In a recent review we defined expertise seeking as the activity of selecting people as sources for consultation about an information need (IPM2014). Such people sources are, however, not in opposition to documentary sources. Rather, we find that engineers look for documents to find people, contact people to be pointed to documents, and interact socially to avoid explicit information searches (IPM2000). Further, engineers seem to turn to people because they often need creative discourse, trusted opinion, and commitment rather than mere facts (ISIC2000). Relative to the importance of people as information sources few systems support people finding. We have suggested personometric mapping as one approach to developing people-finding systems (CoLIS2005).
Source selection. People and documents are established sources of information in the literature, but we also find that experimentation is an important information source (JASIST2021, JASIST2023). Previous work on engineers' information-seeking behavior has led to the formulation of a least-effort principle stating that in choosing their information sources engineers are primarily affected by source accessibility, not source quality. For example, we have found that ease of access has a strong influence on geoscientists' information seeking (JDOC2018). However, we have also found that trust is central to software engineers' choice of information sources. This challenges the least-effort principle by suggesting that engineers' preference for close-by sources can be explained as a choice of sources that are perceived to be more trustworthy, i.e. sources with a higher perceived quality (IWC2002, I&O2002).
Information seeking by novices versus experts. Workplace procedures shape information seeking differently for novices and experts. In a healthcare context we have found that the triage procedure supports less experienced nurses in analytic information seeking, whereas the highly experienced nurses tend to perform triage by holistic pattern recognition. In contrast the timeout procedure facilitates junior as well expert physicians in information seeking by creating a space for open-ended collaborative reflection (IPM2019).
Information seeking by specific groups. We have investigated the information seeking of citizens using e-government (JDOC2022), geoscientists (JDOC2018), international students (JDOC2019), journalists (IPM2022), makerspace users (JASIST2021), physicians and nurses at hospitals (IPM2010, IPM2019), software engineers (ISIC2000, I&O2002), and UX designers (HCII2019, HCII2020).
Computer-supported work
Scientific collaboration. Researchers exploit information technology to collaborate across distance. As a result, many publications are co-authored by researchers from multiple institutions and countries. However, many researchers mostly work alone on their contributions to their joint work. The long stretches of minimal interaction are punctuated by islands of intense interaction (JDOC2019). In a study of social capital in academia, we find that researchers' collaboration preferences do not correlate strongly with their collaboration habits - what they prefer is not necessarily what they practice (RESEVAL2024).
Sustainable fishing. In coastal Indian villages, fishing provides jobs to people who otherwise have few sources of income. A mobile app has been developed to help these fishers locate fish and to warn them about dangerous weather conditions. However, the app also changes the sociomaterial infrastructure by providing free access to information that was previously the privilege of those with experience or expensive equipment (Interactions2018, ICT4D2019). We have also investigated sustainable human-work interaction designs in other contexts (INTERACT2024).
Makerspaces. Makerspaces in public libraries provide possibilities for citizens to make things using technologies such as 3D printers, laser cutters, robotics, and creative materials. We have investigated how librarians scaffold the citizens' learning in library makerspaces (IJCCI2020), how makerspaces afford both individual and collaborative activities (INTERACT2021), and how makers obtain information for their projets (JASIST2021).
Task shifting. Psychiatric medication can have adverse effects such as weight gain, which is a metabolic risk factor. We implemented an IT-supported task shift from physician to pharmacist and found that it improved guideline compliance in the assessment of metabolic risk factors in psychiatric patients. That is, increasing the pharmacist's role in psychiatric care appeared to improve patient outcomes. (IJMI2024).
Students' multitasking. Students use their computers for reading assigned texts, taking notes, and other study activities. However, in class and during preparations for class, they also use their computers for checking social media, for playing online games, and for other distractions. At least, this is the case for Danish students, who perceive such computer multitasking as an integral part of modern life (Interactions2020).
Information visualization
Overview. The notion of overview is focal to information-visualization research. We have reviewed this notion and emphasized the difference between overviews, in the sense of interface components, and overviewing, in the sense of an awareness of an information space (IJHCS2011). It appears to be a frequent, but questionable, assumption that if overviews are provided then overviewing will ensue. Interfaces that aim to support overviewing often provide facilities for filtering the visualized information. We have compared three ways of visualizing the application of filters: blocking, blurring, and color-coding (IJHCS2013). We have also proposed that in collaborative settings information seeking often proceeds as visual overview, oral detail (IJHCS2015). And we have investigated the cost of maintaining an overview in terms of the time spent keeping the overview display current (INTERACT2017).
Fisheye menus. Fisheye menus have become a prominent example of fisheye interfaces, yet they contain several non-fisheye elements and have not been systematically evaluated. We have tried to untangle the impact on usability of various design decisions in fisheye menus. The most striking result is that conventional hierarchical menus appear to be superior to fisheye menus (TOCHI2007).
Visualizing search. In searching a medium-sized collection, we find that browsing is faster than querying. Also, querying based on a Venn diagram interface leads to faster performance than conventional Boolean queries. These differences are obtained without differences in the quality of the participants' solutions of the experimental tasks. However, using a paper copy of the information collection the participants complete tasks faster and reach higher-quality solutions than using the electronic modes of information retrieval (TOCHI1996).
Organizational implementation
Organizational implementation. In brief, organizational implementation is "the process of integrating a technological system into the structures and processes of an organization to accomplish desired change" (M&C2021). It involves the three phases of (a) preparing the implementation, (b) going live with the initial, planned change, and (c) continuing design during use. It is not uncommon for organizations to experience difficulties and delays in organizational implementation (GIQ2016, IJMI2019). In that relation, we have investigated the innovation tactics employed in the work to implement an ML algorithm in healthcare (IJHCS2024).
Design in use. Design does not end prior to use but rather continues, in some form, after systems have entered into operational use. Design in use may be utilized as a deliberate approach to implementation (SJIS2017). We have studied the participatory aspects of such design in use under the rubric of participatory continuing design (PDHIT2017). We have also studied how users perceive a design-in-use approach to implementation (INTERACT2019). To learn about the prerequisites of design in use, we have cataloged the competences needed locally to work with configuring information systems and work practices for each other after go-live (IJHCS2019).
Pilot implementation. An important activity in working with the organizational implementation of information systems is pilot implementation. We have defined pilot implementation as "a field test of a properly engineered, yet unfinished system in its intended environment, using real data, and aiming – through real-use experience – to explore the value of the system, improve or assess its design, and reduce implementation risk" and discussed the challenges involved in conducting pilot implementations (CAIS2012). Learning from pilot implementations is situated, messy, and therefore difficult (HIJ2019). If successful, pilot implementations may bootstrap systems (JCSCW2020). We have also investigated the use of pilot implementations as a means of user involvement in participatory design projects (OZCHI2014), proposed it as a means of evaluating organizational usability (JUS2018), and made a SWOT analysis of pilot implementation on the basis of an INTERACT2021 workshop (Interactions2023).
Digital Post. We have evaluated the adoption and organizational implementation of the the Danish e-government system Digital Post. While it is formally stated that Digital Post will be the major channel for communication between municipalities and citizens, the use of the system was still far from this objective three years after its deployment (ECIS2014). Local government staff are mainly positive about Digital Post at the level of their overall relationship with the system, but at the level of concrete incidents they report a number of situations in which the system decreases the service delivered to citizens (GIQ2016). Relatedly, another study looked at the organizational implementation of document management systems in Danish central government (IIP1995).
Electronic health records. We have evaluated the use of several healthcare systems after they have been in operation for some time. The evaluations differ regarding the amount of use, the user experience of the system, and the organizational implications. A speech-recognition system for making entries in electronic patient records was judged by physicinas to decrease the quality of records and increase the time physicians spent producing the records (IJHCS2009). An e-health system for asthma self-management was not adopted by asthmatics, apparently because the system's image of the user was incompatible with the users' image of themselves (PCEH2009). Additional evaluations of the organizational implementation of healthcare systems appear in the section on health informatics, see below.
Health informatics
Emergency-department whiteboards. Creating and maintaining an overview is important to hospital clinicians especially at wards that treat acute patients. At emergency departments (EDs) whiteboards are a central tool for supporting clinicians in creating and maintaining the overview they need in their work. We have conducted a field study of the use and role of dry-erase whiteboards at EDs (JCSCW2011), field and interview studies of the organizational implementation of an electronic whiteboard at four EDs (IRIS2010, PDC2012), a survey of clinicians' expectations toward electronic ED whiteboards (IJMI2011), and before/after measurements of the work-practice changes associated with the introduction and use of an electronic ED whiteboard (MIE2012, HIJ2013, HIJ2016). We also investigated the coordinative unit consisting of the whiteboard and the coordinating nurse (IJHCS2015), the effect of the electronic whiteboard on the noise level at the EDs (EOJ2013), and the possibility of using the whiteboard data to visualize and counteract crowding (EOJ2016, EOJ2017). In spite of the success of the whiteboard in the EDs the whiteboard data are incomplete to the extent of preventing most types of reuse (MEDINFO2017).
Electronic medication record. System adoption is often delayed or hampered by barriers. We have evaluated the gap between actual and mandated use of an electronic medication reord after it has been used for 1.5 to 4 years (MIE2008, EJISE2012). Reasons for the identified gap has been discused in terms of breakdowns in collaborative informative seeking (IPM2010). In an action-research study we followed up on the gap by performing interventions to improve the actual use of the electronic medication record at a selected ward and by measuring the effect of the interventions (GROUP2009). We have also attempted to map out the infrastructural complexity of the medication process across the different healthcare sectors (InfraHealth2023) and showed how electronic medication management involves continual repair work (MEDINFO2023).
Implementating Epic's electronic health record. We have compared the experiences from implementing Epic in Denmark and Finland (IJMI2022) and compared the Norwegian preparations for Epic with the experiences from implementing it in Denmark and the UK (IJMI2019). In the Region of Central Norway, Epic is being implemented at the hospitals and also offered for use in general-practitioner clinics (HIJ2021, JCSCW2022) and in municipal healthcare (CSHI2021, MIE2022). At the hospitals, the secretaries have had problems getting the system configured for their workflows (CSHI2023). In addition, the system obviously has consequences for the suppliers of the existing Norwegian EHRs (InfraHealth2021).
Other electronic patient records. We evaluated the effects of the clinical-process module of an electronic patient record on three clinical activities at a stroke unit. Compared to the use of paper records, the clinical-process module, for example, reduced physicians' mental workload (IJMI2008). Also, a study of icons that showed the status of blood tests concluded that the icons passed by most of the complexity of the blood-test process (MEDINFO2013) and mostly remained unseen by the physicians (CSCW2015). We have also studied the implementatiion of open notes in Sweden (MIE2024) and engaged in a participatory-design process to provide IT support for the patient handover from intensive care unit to general ward (MEDINFO2019). Relatedly, we have organized a participatory-design effort to devise better IT support for the coordination of surgical operations (JCSCW2020) and investigated how different digital trajectories were brought into convergence to evolve the information infrastructure at a hospital (JCSCW-forthcoming). Studies of several other healthcare systems appear above.
Long-term care. Apart from relatives, the two main providers of long-term care are home-care services and nursing homes. Home care is an example of a highly mobile work setting. We have investigated the role of rhythms in home-care work as well as in individual home-care workers' command of their work (GROUP2005). We have also given a countrywide status of the documentation-related risks to patient safety in long-term care in Denmark. The status is that care, medication, and consent are inadequately documented (ACI2021). To enrich welfare technologies, we have proposed to make them more collaborative by means of me-to-we design (MIE2023).
Effects-driven IT development and improvement
Effects-driven IT development. We propose to supplement functional specifications with specifications of the effects desired from using a system and assessments of whether these effects are achieved. The aim of effects-driven IT development is to provide an instrument for managing IT projects by making it possible to work systematically toward developing efficient solutions that meet customer goals. We have described effects-driven IT development (SJIS2011) and summarized our experiences from our empirical projects (BSIISD2011). A central element of effects-driven IT development is a sustained focus on usage effects, which extends an iterative, participatory-design approach into the implementation of IT systems (DI2012, PDC2012). While we consider effects-driven IT development applicable to a wide range of IT projects, we have mainly investigated it within the healthcare domain (NordiCHI2004, SHI2005, PDC2006, PDC2008, PDC2010, CUEDDP2013).
Effects-driven IT improvement. Many initiatives to optimize IT use happen long after a system has been developed and gone live. In our experience, IT improvement is often a local initiative that exploits a brief window of opportunity, is conducted in a single department, enjoys little attention from top management, and makes do with few resources. We propose an effects-driven approach to such improvement initiatives (SJIS2022). This apprach is based on a series of action research studies (e.g., GROUP2009, EJISE2012, HIJ2013, HIJ2016, PDHIT2017, JCSCW2020, MIE2020). We have also proposed that our effects-driven approach can be used as a model for quality development in healthcare (NJWLS2018).
Pilot implementation. An activity central to effects-driven IT development is pilot implementation (CAIS2012), see above.
Document management and information retrieval
Classification and its consequences. In managing recorded information such as documents we need access points into the collections. A study of access points into a collection of legal texts shows the need for flexible classifications, which on the one hand provide an ordering shared by all users and on the other hand are adaptable to the specific needs of individual users and to changes that occur over time (EJIS1993). Another study analyzed how a classification scheme for managing the requirements for a software system was brought about and how it entered into constituting the requirements specification process (JCSCW2004).
Managing and using documents. To design document management systems we need to understand how documents help people accomplish their tasks. For this purpose we have delineated six roles documents play in professional work (ECSCW1999). Our basis for identifying the six document roles were, among other things, our previous studies of document management practices: a study of how biochemists in an industrial R&D lab manage the documents they keep in their offices (IRIS1993) and a study of how Danish central government institutions manage their documents (IIP1995).
Retrieval from structured documents. We have investigated how people intertwine searching and reading during retrieval of information from hierarchically structured software manuals (INTERACT2001). This work is to help us understand when a cluster of text chunks that match a query should be returned as individual hits and when it is preferable to return a single, superordinate chunk that provides an entry point to the entire cluster. The work on best entry points for structured document retrieval has been continued in relation to a collection of Shakespeare plays (IPM2006a, IPM2006b).
Multimedia information retrieval. On the basis of a number of information requests received by a large film archive we have analyzed what types of information needs real users have and how these needs are expressed. Findings include that the attributes used in specifying information needs relate to the production, content, subject, context, and screening of films (JDOC2003). We have also investigated how people ask questions about music on social Q&A sites (JDOC2017) and started investigating whether speech recognition in Danish is sufficiently accurate for retrieving radio broadcasts (OZCHI2016).
Information systems development
Conceptual design. One study investigated how a small-scale classification scheme for managing system requirements was brought about and how it entered into constituting the requirements-specification process (JCSCW2004). Another study analyzed how scenarios attained a central role in a conceptual design by providing a descriptive, task-based account of the users' work (IJHCS2003). In terms of providing tools for designers, we have sketched the design of a low-tech tool for supporting the generation and management of an inclusive set of system requirements (TMCE2000).
User-centered design. The complexities of software design have consequences for user-centered design. These implications concern the articulation of a system design, the pressure to produce progress, and the challenges of collaboration (EIS2007). To work systematically with the articulation and organizational implementation of systems, we have proposed a sustained participatory-design approach (PDC2008) and are devising effects-driven IT development. Recently, we have provided an overview of the research on participatory design (HCIUEMT forthcoming). We have also investigated how UX designers make decisions (HCII2020). A specific element of participatory design concerns how users are selected for participation. We have investigated whether representative samples of users are traded for advocates and champions (IWC2011). Years ago, a review of museum web sites suggested a lack of user-centered design (ARMU1998).
Configuration and implementation of large-scale suite systems. We have investigated the introduction of the Epic electronic health record in Denmark and Norway (IJMI2019). In Norway we have also investigated the user participation during the preparations for the introduction of Epic (InfraHealth2019, MIE2020). Specifically, we have analyzed the drivers of the general practitioners' expectations toward Epic (HIJ2021).
Outsourcing. The distribution of development activities across multiple development organizations has become widespread. We have studied Danish financial organizations' offshore outsourcing of information systems development to Indian development organizations (BSIISD2011a). In one of these studies a minimal-interaction strategy was seen as a solution to the cultural and maturity inequality between the client and vendor organizations (BSIISD2011b).
Situated methods for design. We have investigated design as a situated activity and described a series of situated design methods (SDM2014a), one of them specifically aimed at supporting design students in focusing their design projects (SDM2014b). In continuation of this work, we have started to devise a framework for reflecting on design processes (OZCHI2014).
Pointing with mouse and touchpad
Target precuing. In point-and-click interfaces the location of targets is sometimes known to the user before visually identifying it, and sometimes not. We find that target precuing affects trial completion time, reaction time, and multiple movement kinematics and that it affects pointing with the mouse and touchpad differently (IJHCI2013). These results, for example, complicate cross-study comparison of pointing performance because the tasks used in such studies often differ considerably in the precues they provide.
Age effects on pointing performance. We have investigated young, adult, and elderly participants’ performance when pointing with mouse and touchpad (IJHCI2010). Adult participants performed better than both young and elderly participants in that adult participants made fewer errors than young participants and completed trials quicker than elderly participants. Moreover, young participants were quicker than elderly participants. All three age groups were slower and made more errors with the touchpad than the mouse.
Cell cursors. Touchpad devices are widely used but slower than pointing with a mouse. The TouchGrid, an instance of what we term cell cursors, replaces moving the cursor through dragging the finger on a touchpad with tapping in different regions of the touchpad. The touchpad regions are recursively mapped to smaller display regions and thereby enable high-precision pointing without requiring high tapping precision (BIT2005). Two further variants of cell cursors - box and pie cursors - have been investigated as examples of cursors that dynamically change their cursor activation area (IJHCS2007).