Contributions published at Human Aspects of Software Engineering (Thomas Fritz)

The modern workplace demands more of its employees than ever before as technology assumes a larger role. In this context, the productivity and wellbeing of the individual as well as the team and the overall organizations is increasingly important, yet more research is needed to understand them and their relation. In a set of two research studies, we explored the productivity and wellbeing of software developers in team dynamics. We first examined the effects of virtual and traditional work environments on knowledge workers while performing work-related tasks. The qualitative and exploratory quantitative results show that the closed office and the beach VR are similarly good according to users’ ranking in reducing distractions and inducing flow, and that these two environments are preferred over the non-VR open office and VR open office environments. Overall, these results indicate the potential that VR environments have to help knowledge workers achieve flow and stay calm and focused even in loud open office work settings. Second, we designed and piloted a multi-month field study in which participants regularly respond to a series of productivity and wellbeing questionnaires throughout their workday. The multi-tiered data collection includes hourly, daily, weekly, monthly, and one-time surveys for a better understanding of the fine- and coarse-grained factors which affect the productivity and wellbeing of software teams. The broader vision of our research is to better understand productivity and wellbeing on an individual and team level, and to develop approaches that support professionals in spending their time well at work. By taking a more holistic approach and incorporating developer wellbeing, we can design more effective strategies for an upwards cycle of productivity and wellbeing.

Knowledge workers face many challenges in the workplace: work is fragmented, disruptions are constant, tasks are complex, and work hours can be long. These challenges can affect knowledge workers’ stress, focus and awakeness, and in turn their interaction with the digital environment, the quality of work performed and their productivity in general. We report on a field study with 14 knowledge workers over an eight-week period in which we investigated, using experience sampling, how the workers experience stress and awakeness over time. During this field study, we also collected biometric data including heart- and skin-related measures, which we then used to investigate if it is possible to predict stress, focus and awakeness, in the moment. We observed and report on various trends in knowledge worker stress and awakeness levels over several weeks, finding that people tend to have certain “baseline” levels for these aspects. Moreover, we found that days with high levels of stress tend to cluster, similarly as the days with low awakeness. We further show that machine learning models can be built from the data of a single minimally invasive device to predict stress, focus, and awakeness. Overall, we found that our models were capable of large improvements in precision and recall in comparison to a random classifier for stress (25.9% increase over random for precision, 4.2% for recall) and awakeness (52.4% increase in precision, 40.8% in recall). The abstract concept of focus proved to be the hardest to predict (26.0% increase in precision, 27.8% decrease in recall).

A software developer works on many tasks per day, frequently switching between these tasks back and forth. This constant churn of tasks makes it difficult for a developer to know the specifics of when they worked on what task, complicating task resumption, planning, retrospection, and reporting activities. In a first step towards an automated aid to this issue, we introduce a new approach to help identify the topic of work during an information seeking task — one of the most common types of tasks that software developers face — that is based on capturing the contents of the developer’s active window at regular intervals and creating a vector representation of key information the developer viewed. To evaluate our approach, we created a data set with multiple developers working on the same set of six information seeking tasks that we also make available for other researchers to investigate similar approaches. Our analysis shows that our approach enables: 1) segments of a developer’s work to be automatically associated with a task from a known set of tasks with average accuracy of 70.6%, and 2) a word cloud describing a segment of work that a developer can use to recognize a task with average accuracy of 67.9%.

Knowledge workers receive multiple inquiries via communication messages, often interfering with their current workflow and leading to context switches, consequently impeding progress on the endeavored task. To address this quandary, we investigate the following research question: Can we develop an approach that 1. assists the participant to focus on a specific task and 2. reduces the number of context switches during this timeframe? This thesis establishes a particular set of concepts and describes their implementation in a prototype called FocusSession. In a subsequent preliminary evaluation, as outlined in this research, participants examine this prototype in their respective work environments. Interpreting the data and interview transcripts, FocusSession appears to positively impact the participants' focus continuation on a specific task while enhancing the perception of context switches. Furthermore, subsidiary areas of interest for future research are drafted, such as establishing different profiles to cater to knowledge workers' diverse needs. We acknowledge the necessity for a study with adequate sample size to validate or further investigate the preliminary evaluation's findings. We are confident that the FocusSession approach helps knowledge workers be more productive overall by empowering them to focus on a specific task for a selected timeframe.

Modern knowledge work is increasingly reliant on online resources. Web browsers are, however, not optimized for the task-based workflows and information gathering needs of knowledge workers. We propose an approach that supports users in curating groups of browser tabs that belong to a task, helping them with task switching and resumption. Furthermore, based on users’ browser behavior, we automatically derive such groups and propose them for addition. An in situ user study showed that users like the approach and would continue using such a system, but that there are challenges ahead in providing relevant and reusable suggestions.

The modern workplace is more demanding than ever before. Yet, since the industrial age, productivity measures have predominantly stayed narrowly focused on the output of the work, and not accounted for the big shift in the cognitive demands placed on the workers or the interleaving of work and life that is so common today. We posit that a more holistic conceptualization of Time Well Spent (TWS) at work could mitigate this issue. In our 1-week study, 40 knowledge workers used the experience sampling method (ESM) to rate their TWS and then define TWS at the end of the week. Our work contributes a preliminary characterization of TWS and empirical evidence that this term can capture a more holistic notion of work that also includes the worker’s feelings and well-being.

Software developers, like other information workers, continuously switch tasks and applications to complete their work on their computer. Given the high fragmentation and complexity of their work, staying focused on the relevant pieces of information can become quite challenging in today’s windowbased environments, especially with the ever increasing monitor screen-size. To support developers in staying focused, we conducted a formative study with 18 professionals in which we examined their computer based and eye-gaze interaction with the window environment and devised a relevance model of open windows. Based on the results, we developed a prototype to dim irrelevant windows and reduce distractions, and evaluated it in a user study. Our results indicate that our model was able to predict relevant open windows with high accuracy and participants felt that integrating visual prominence into the desktop environment reduces clutter and distraction, which results in reduced window switching and an increase in focus.

Open offices are cost-effective and continue to be popular. However, research shows that these environments, brimming with distractions and sensory overload, frequently hamper productivity. Our research investigates the use of virtual reality (VR) to mitigate distractions in an open office setting and improve one’s ability to be in flow. In a lab study, 35 participants performed visual programming tasks in four combinations of physical (open or closed office) and virtual environments(beach or virtual office). While participants both preferred and were in flow more in a closed office without VR, in an open office, the VR environments outperformed the no VR condition in all measures of flow, performance, and preference. Especially considering the recent rapid advancements in VR, our findings illustrate the potential VR has to improve flow and satisfaction in open offices.

The goal of this thesis is to evaluate whether it is possible to detect interruptions of knowledge workers in a open-plan office using audio and video recordings. To achieve a proof of concept, two separate pilot studies were conducted at an industrial company in November and December of 2018. In the first study, two cameras were installed such that they could produce a stereoscopic representation of the surroundings. People were detected on the video recording by applying the image differencing technique known in computer vision. Through triangulation of the detected people, their position in 3D space could be determined. This information was then processed in a heuristic which aims at recognizing interactions between people and interruptions in the work place based on their distance in space and their movements. In the second study, the pre-existing self-monitoring software PersonalAnalytics was amended by a new module which stores the audio signal received from a connected omni-directional microphone to a local file and tries to determine whether someone is speaking. For this purpose, a software for speaker diarization was used which uses mel-frequency cepstrum coefficients as features and applies a Gaussian mixture model on them. The data gathered in these studies were analysed and the quality of the detection was evaluated. This showed that the approach to detect interruptions using audio and video data is practicable in principle but needs still some improvement to compromise a useful system. This applies in particular to the method used for voice activity detection which produces an intolerable high amount of false positives. Based on the findings of this study, additional research is needed to cope with aforementioned challenges. First, improvements to the algorithms used in this approach are possible and should be implemented. Second, the approaches for interruption detection could be combined to achieve a more reliable system.

Software development organizations strive to enhance the productivity of their developers. All too often, efforts aimed at improving developer productivity are undertaken without knowledge about how developers spend their time at work and how it influences their own productivity. In our research, we focus on two aspects for improving developers' productivity: better understanding developer productivity and using these findings to foster productivity at work. To better understand developer productivity, we took a bottom-up approach by investigating developers' perceptions of productivity in the field, and by examining the individual differences in each developer's work. We found that developers spend their time on a wide variety of activities and tasks that they regularly switch between, resulting in highly fragmented work. Extending our understanding of developers' work and the factors that impact their productivity then allowed us to develop models of developers' work and productivity, and build approaches that support developers with productive behavior changes. To support the identification of self-improvement opportunities that motivate productive behavior changes, we studied how we can increase developers' awareness about work and productivity by combining our models with three persuasive strategies: self-monitoring, self-reflection, and an external indicator. Based on successful applications in the health and physical activity domain and from examining developers’ expectations, we developed PersonalAnalytics, a workplace self-monitoring tool that collects a broad variety of computer interaction data and summarizes the data in a daily and weekly retrospection. A multi-week field-study showed that PersonalAnalytics offered meaningful insights to 82% of the participants, but the insights were not actionable enough to motivate behavior change for 41% of our participants. In a follow-up study, we found that continuous and purposeful self-reflection can motivate productive self-improvements in the workplace, since 83% of our participants stated that it supported the identification of goals and actionable strategies, and 80% reported productivity increasing behavior changes. We further studied how we can increase developers' awareness about their co-workers' availability for interruptions, by sensing and externally indicating interruptibility to developers based on their computer interaction. Our large-scale field study with the FlowLight showed that we can effectively reduce 46% of external interruptions, participants felt more productive, and 86% of them remained active users even after the two-month study period ended. Overall, our research showed that we can successfully foster productivity at developers' work, by increasing their awareness about productive and unproductive work habits, and by encouraging work habit improvements based on the gained insights. In addition, our research can be extended and opens new opportunities to foster productive work for development teams.

Emotions have an impact on many aspects of our lives like productivity, well-being, learning, an many more. The ability for computing systems and applications to recognize, interpret, and adapt to these emotions has gained significant importance over the past decades and therefore, the multidisciplinary research field Affective Computing emerged. Tools developed in the do- main try to recognize emotions in real time, provide affect awareness with visualizations, and support the users to interpret and regulate emotions and emotion-motivated behavior with feed- back interventions. This thesis is focusing on supporting specifically knowledge workers with the development of such an Affective System that increases their affect awareness and intervenes when their mood gets negative. This Affective System was implemented by extending the existing personal infor- mation system PersonalAnalytics. We were able to show that it is possible to recognize emotions via facial expression analysis in a minimally invasive way. We further designed and developed multiple visualizations with different levels of abstraction to provide knowledge workers with the best overview of emotions and how they connect to their tasks. Finally, affective feedback interventions were developed, with the aim to increase their mood towards the positive. To do that, a set of interventions triggered based on different rules was created. We were able to show that these interventions increased the mood. With this affective extension of PersonalAnalyt- ics, we can support knowledge workers to reflect on, interpret and regulate their emotions and emotion-motivated behavior.

Meetings are a big part of work life. While they are important, meeting participants often perceive them as a waste of time. Attending bad meetings can not only negatively affect the general mood of employees but also lower their job satisfaction. Researchers have looked into an automatic analysis of meeting quality based on features that can be extracted from the speech. However, all of these studies are limited as they often depend on the human coding of events from recorded audio data or on usage of scenario meetings in which each participant is given a specific role. This thesis explores whether it is possible to measure participants' perceived quality of meetings using automatically extractable features of audio recordings from non-scenario meetings. To investigate this, we conducted a multi-day field study in a company based in Germany. The gathered data consists of 25 raw audio recordings from the meetings as well as 78 answers of the post-meeting survey that assesses the perception of meeting quality. We developed an approach to extract speech-related features of a raw audio file which we used to analyze the collected audio recordings. The results of our survey indicate that an open and positive meeting atmosphere and a lively exchange in meetings are both important factors contributing to participants' meeting quality. Results further suggest that the number of speaking turns is the only factor that we captured automatically and that is related to the meeting quality. Nevertheless, we see a potential to increase the meeting quality in the future, either by providing awareness about participant's meeting behavior or by reflecting on the meeting quality over time.

null

Interruptions at the workplace can consume a lot of time and cause frustration, especially if they happen at moments of high focus. To reduce costly interruptions, we developed the FlowLight, a small LED Lamp mounted at a worker's desk that computes a worker's availability for interruptions based on computer interaction and indicates it to her coworkers with colors, similar to a traffic light. In a large study with 449 participants, we found that the FlowLight reduced interruptions by 46%. We also observed an increased awareness of the potential harm of interruptions and an increased feeling of productivity. In this chapter, we present our insights from developing and evaluating FlowLight, and reflect on the key factors that contributed to its success.

To overcome the ever-growing demand for software, software development organizations strive to enhance the productivity of their developers. But what does productivity mean in the context of software development? A substantial amount of work on developer productivity has been undertaken over the past four decades. The majority of this work considered productivity from a top-down perspective (the manager view) in terms of the artifacts and code created per unit of time. Common examples of such productivity measures are the lines of source code modified per hour, the resolution time for modification requests, or function points created per month. These productivity measures focus on a single, output-oriented factor for quantifying productivity, and do not take into account developers’ individual work roles, practices and other factors that might affect their productivity, such as work fragmentation, the tools used, or the work/office environment. In our research, we investigated how productivity could be quantified from the bottom-up, following a mixed-methods approach that involved more than 800 software developers. By investigating developers’ individual productivity, it is possible to better understand the individual work habits and patterns, how they relate to the productivity perceptions and also which factors are most relevant for a developer’s productivity.

Recently, we have seen an explosion in the number of devices and apps that we can use to track various aspects of our lives, such as the steps we walk, the quality of our sleep, or the calories we consume. People use devices such as the Fitbit activity tracker to increase and maintain their physical activity level by tracking their behavior, setting goals (e.g. 10'000 steps a day) and competing with friends. Many of these approaches have been shown to successfully encourage users to change their behaviors, often motivated through persuasive technologies, such as goal-setting, social encouragement and sharing mechanisms. We explored how we can map the tremendous success of these smart devices to the workplace, with the aim to increase software developers' self-awareness about productivity through self-monitoring. Yet, little is known about expectations of, the experience with, and the impact of self-monitoring in the workplace. From a mixed-methods approach we inferred design elements for building workplace self-monitoring tools, which we then implemented as a technology probe called WorkAnalytics. We field-tested these design elements during a three-week study with software development professionals. In the field study, we found that self-monitoring paired with experience sampling increases developers' awareness about work and motivates many to improve their behaviors, and that a wide variety of different metrics is needed to fulfill developers' expectations. Our work can serve as a starting point for researchers and practitioners to build self-monitoring tools for the workplace.

Software development is a complex task and developers need to use a variety of applications for their daily work. The demand to transfer information makes it necessary to use many of the applications simultaneously. This often results in a myriad of open windows, which may reduce navigation efficiency and focus as the number of windows increases. In this thesis, we analyzed how professional software developers interact with and manage their desktop environment. In particular, we look to see if developer efficiency could be affected by high numbers of opened windows. To construct the dataset, we observed the desktop environments of 12 professional software developers from three different companies over a combined total of 195 days. For this task, we used a monitoring application also capturing visual focus with an eye tracker. The eye tracker provided valuable insights additional to the traditional interaction data. We found only 79\% of the visual attention was directed at the window with the keyboard input. Half of the desktop environments had 10 or more windows open while mostly only two were fully visible. The number of open windows grows over the course of a work day and we learned most developers do not proactively close windows. From time to time, we could observe desktop environments go through cleanup cycles. Nonetheless, found evidence of unused windows overcrowding the desktop environments and see potential to foster developers focus in the future.

Building and maintaining modern software systems requires developers to perform a variety of tasks that span various tools and information sources. The crosscutting nature of these development tasks requires developers to maintain complex mental models and forces them (a) to manually split their high-level tasks into low-level commands that are supported by the various tools, and (b) to (re)establish their current context in each tool. In this paper we present Devy, a Conversational Developer Assistant (CDA) that enables developers to focus on their high-level development tasks. Devy reduces the number of manual, often complex, low-level commands that developers need to perform, freeing them to focus on their high-level tasks. Specifically, Devy infers high-level intent from developer's voice commands and combines this with an automatically-generated context model to determine appropriate workflows for invoking low-level tool actions; where needed, Devy can also prompt the developer for additional information. Through a mixed methods evaluation with 21 industrial developers, we found that Devy provided an intuitive interface that was able to support many development tasks while helping developers stay focused within their development environment. While industrial developers were largely supportive of the automation Devy enabled, they also provided insights into several other tasks and workflows CDAs could support to enable them to better focus on the important parts of their development tasks.

Software development organizations strive to enhance the productivity of their developers. While research has looked into various ways for improving developer productivity, little is known about the activities they pursue at work, how these activities influence the fragmentation of work, and how these insights could be leveraged to foster productivity at work. In my PhD thesis, I address software developer productivity by taking a mixed-method approach to investigate developers’ perceptions of productivity in the field and to examine the individual differences of each developer’s work. My goal is to increase developers’ awareness about their own work habits and productivity, and to encourage productive behavior changes at work through the provision of two persuasive technologies, self-monitoring and goal-setting.