Linköpings universitet
Linköping University | Department of Computer and Information Science
Bachelor’s thesis, 18 ECTS | Cognitive Science
2020 | LIU-IDA/KOGVET-G--20/009--SE
We Move in Order to Perceive
–
A Mouse-tracking Study of User Behaviour During Stalling
Branched Videos with a Playback Bar
Ebba Fogelberg
Supervisor: Erkin Asutay Examiner: Kenny Skagerlund
Upphovsrätt
Detta dokument hålls tillgängligt på Internet - eller dess framtida ersättare - under 25 år från publicer-ingsdatum under förutsättning att inga extraordinära omständigheter uppstår.
Tillgång till dokumentet innebär tillstånd för var och en att läsa, ladda ner, skriva ut enstaka ko-pior för enskilt bruk och att använda det oförändrat för ickekommersiell forskning och för undervis-ning. Överföring av upphovsrätten vid en senare tidpunkt kan inte upphäva detta tillstånd. All annan användning av dokumentet kräver upphovsmannens medgivande. För att garantera äktheten, säker-heten och tillgängligsäker-heten finns lösningar av teknisk och administrativ art.
Upphovsmannens ideella rätt innefattar rätt att bli nämnd som upphovsman i den omfattning som god sed kräver vid användning av dokumentet på ovan beskrivna sätt samt skydd mot att dokumentet ändras eller presenteras i sådan form eller i sådant sammanhang som är kränkande för upphovsman-nens litterära eller konstnärliga anseende eller egenart.
För ytterligare information om Linköping University Electronic Press se förlagets hemsida http://www.ep.liu.se/.
Copyright
The publishers will keep this document online on the Internet - or its possible replacement - for a period of 25 years starting from the date of publication barring exceptional circumstances.
The online availability of the document implies permanent permission for anyone to read, to down-load, or to print out single copies for his/hers own use and to use it unchanged for non-commercial research and educational purpose. Subsequent transfers of copyright cannot revoke this permission. All other uses of the document are conditional upon the consent of the copyright owner. The publisher has taken technical and administrative measures to assure authenticity, security and accessibility.
According to intellectual property law the author has the right to be mentioned when his/her work is accessed as described above and to be protected against infringement.
For additional information about the Linköping University Electronic Press and its procedures for publication and for assurance of document integrity, please refer to its www home page: http://www.ep.liu.se/.
Abstract
This thesis analyses how users’ mouse behaviour during a video stall gets influenced by the type of video, either branched or linear, and by the presence of a playback bar. An experiment was conducted with thirty-two participants divided into six groups. Each group was watching a different combination of four videos with stalls, the first two videos belonging to the same type of video and either with or without a playback bar, and the last two videos changed in one of the two aspects. With mouse-tracking, these aspects were studied through the variables of mouse activity, average speed, average distance between the cursor and the playback bar, and the total distance moved on the screen. The partici-pants also filled in questionnaires about their mouse behaviour, after watching each video, and their answers were later analysed through a thematic analysis.
The results showed no significant differences between the groups in any of the main dependent variables. In general, within all groups, the participants moved the mouse very scarcely, indicating that the results about mouse movement should be interpreted carefully. During videos with a playback bar, mouse movements appeared to be concentrated to the stalls, focusing the movements to the bottom of the screen where the playback bar is located. Mouse behaviour during videos without a playback bar was more evenly divided between the different parts of the video and of the screen, or the user were not moving the mouse at all. Watching branched or linear videos influenced the mouse behaviour in such a way that branched videos seemed to engage the user to interact with the video player more than the linear videos. However, no difference was noticed between these conditions for active users during stalls when a playback bar was present. The thematic analysis gave clear indications that the playback bar was an important component for understanding a stall.
Based on these findings, conclusions are drawn that a stall is a situation of watching videos during which mouse behaviour may be less affected by the type of video, and more influenced by the access to a playback bar. The playback bar was shown to be a source of information about the system and the situation.
Keywords: branched video, interactive video, linear video, playback bar, mouse be-haviour, mouse-tracking, video stall, user interaction.
Acknowledgments
I want to express my gratitude to my supervisor Erkin Asutay for great guidance, and to Niklas Carlsson for all the help and for introducing me to this interesting subject. I also want to give a big thank you to Eric Lindskog who has responded with great patience to all my questions about the video player and the system used in the experiment. I am thankful to all participants who contributed their time by taking part in the experiment, and to Peter Fogelberg and Frida Heskebeck for proof reading and giving helpful comments.
Last but not least, I want to express my gratitude to Wilhelm Brodin for great teamwork and support during these months. And for answering all my silly questions.
Linköping, June 2020 Ebba Fogelberg
Contents
Abstract iii
Acknowledgments iv
Contents v
List of Figures vii
List of Tables viii
1 Introduction 1
1.1 Aim and Research Questions . . . 1
1.2 Delimitations . . . 2
1.3 Thesis Outline . . . 2
2 Background 3 2.1 Supporting the User . . . 3
2.2 Branched Video . . . 4
2.3 Video Stall . . . 6
2.4 Playback Bar . . . 7
2.5 Generic Video Player . . . 8
2.6 Mouse Behaviour . . . 8 3 Method 10 3.1 Experimental Design . . . 10 3.2 Participants . . . 11 3.3 Equipment . . . 12 3.3.1 Setup . . . 12 3.3.2 Video Player . . . 13
3.3.3 Videos and Stalls . . . 13
3.3.4 Mouse-tracking . . . 14
3.4 Procedure . . . 15
3.5 Pilot Studies . . . 16
3.6 Analysis . . . 16
4 Results 17 4.1 Difference Between Having a Playback Bar or Not . . . 17
4.1.1 Playback Bar During Branched Videos . . . 17
4.1.2 Playback Bar During Linear Videos . . . 18
4.1.3 Order Effects of Switching Between Conditions . . . 18
4.2 Difference Between Branched and Linear Videos . . . 19
4.3 Comparing Stall with the Whole Video . . . 20
5 Discussion 23
5.1 The Effect of Having a Playback Bar . . . 23 5.2 The Effect of Watching a Branched Video . . . 26 5.3 Method Discussion . . . 27
6 Conclusion 30
List of Figures
2.1 Example structure of a branched video. Grey circles symbolise start and end nodes; white circles symbolise branch points; the red line symbolises that only one path may be chosen. . . 4 2.2 Display of a branch selection in the video player used in this study. Rabbit hole and
Happy bunny are the two alternatives that the user may choose between. . . 5 2.3 Playback bar for linear videos. . . 7 2.4 Playback bar for branched videos. . . 7 3.1 The four combinations of type of video and video player studied in this thesis. . . 10 3.2 The six groups, based on three aspects, that the participants were divided into.
Playback bar is abbreviated to pbb. . . 11 3.3 The participants’ weekly habits of watching videos and playing computer games. 12 3.4 Graphic representation of the structure of video 1 . . . 14 3.5 Graphic representation of the structure of video 2-4 . . . 14 4.1 Distribution of mouse activity during stall (red), branch selection (blue), and
dur-ing regular playback (grey), for groups BPBN and BNBP. . . 20 4.2 Distribution of mouse activity during stall (red), branch selection (blue), and
dur-ing regular playback (grey), for groups BPLP and LPBP. . . 20 4.3 Distribution of mouse activity during stall (red), and during regular playback
List of Tables
1
Introduction
Watching videos is something a lot of people do on a weekly basis. Streaming services supply users with a large range of genres whenever the users desire to watch a video. Having these various streaming services enables users to more easily consume multiple videos and makes mass production of videos more viable. However, most (if not all) streaming services today are designed for non-interactive linear videos, lacking generic ways to display other types of videos, such as interactive branched video that will be studied in this thesis.
Interactive branched video is a type of video where the viewer can influence the story that is told. Whilst the video is playing, the viewer gets to choose between alternative plot lines, reaching different ends depending on their choices. This is a type of interactive storytelling and can be compared to Choose Your Own Adventure books. Branched video is different from the typical kind of video, here referred as a non-interactive linear video, or just linear video. It is non-interactive because the viewer cannot influence the video, and linear because it only contains one story line, always showing the same content, from start to finish.
Although most videos today are linear, branched videos have been released on various streaming platforms, although most often referred to as simply interactive, not using the term branched. The streaming service Netflix has released a couple of interactive videos, with the most well-known being Black Mirror: Bandersnatch from 2018, which they refer to as an "in-teractive odyssey with multiple endings" [1]. However, the interactivity is designed and implemented differently on separate platforms, and sometimes even within the same plat-form. In these cases, the videos are released and presented without a generic video player or interface. In comparison with linear videos, presenting a branched video demands spe-cific implementation, due to the added functionality of exposing choices to the viewer [2]. Creating a generic player for branched videos could contribute to making this medium grow and be more accessible, since the framework would handle concerns regarding the video pre-sentation, exposing the choices to the viewer, and keeping track of how the different video segments may be connected.
Previous research within this area has mainly focused on the system behind the player’s interface, for example when and how the player should load the upcoming videos [3]. Recent research, however, highlights the importance of creating a generic interface for this type of interactive video, focusing on how specific components should be designed to give the viewer the best experience [4]. To ensure that the video player induces good user experience, it is important to study and evaluate how the user interacts with the interface and its elements during different events. It has been suggested that the specific situation of a stall, i.e. when the video suddenly freezes, may bring forth different user needs and user behaviour in relation to a playback bar [4]. Therefore, this thesis focuses on how the user interacts with the interface during stalls, comparing videos with a playback bar to videos without one.
1.1
Aim and Research Questions
The underlying purpose of this thesis is to understand how a user interacts with the video player’s interface during a stall. To reach this understanding, the aim of this thesis is to
1.2. Delimitations
identify if, and in that case how, user behaviour during stalls differ depending on the type of video and whether the user has access to a playback bar.
User interaction will be studied through the users’ mouse behaviour, which in turn will be gathered through mouse-tracking, recording cursor movement. This will be accompanied with the users’ self-reported descriptions of their mouse behaviour. The following research questions will be studied:
1. How does user behaviour during stalls differ if the user has access to a playback bar or not?
2. How does user behaviour during stalls differ if the user is watching a branched or a linear video?
1.2
Delimitations
Interactive videos can take form in many different ways. In this thesis, the only type of interactive video included is videos that are interactive in the sense that they are branched and the viewer chooses which plot line to follow. Interactive videos may be linear, but no such videos are included in this project; the linear videos used in this experiment are non-interactive. Branched videos may also be non-interactive, but this type of video is not covered in this thesis either.
The type of user behaviour studied in this thesis is limited to the users’ mouse movements and their self-reported experience of watching videos and strategies of interpreting the stalls. Mouse movements will be recorded through cursor movement on the screen, stated in pixels. Other measurements, such as pressure of the finger during a mouse click and wrist move-ments, are not recorded due to lack of equipment as well as prioritising mouse-tracking to be hidden from the participants during the experiment.
1.3
Thesis Outline
This first chapter has introduced the topic and presented the purpose and delimitations of this thesis. Following is Chapter 2 with information about different types of videos, explaining terms such as branched video, playback bar and stall. It also contains information from previous research about human-computer interaction and using mouse-tracking as a method and what type of results it may give.
Chapter 3 presents the used method, the experimental design, how the experiment was conducted, the participants, and how the analysis was performed. In Chapter 4, the results from the experiment are presented, divided into subsections for differences due to having a playback bar or not, in branched and linear videos separately, or differences due to the type of video. Then the thematic analysis of the participants’ self-reported answers about their mouse behaviour is presented.
Chapter 5 is where the results and the method are discussed in relation to previous re-search. Finally, Chapter 6 contains the conclusion of this thesis.
2
Background
This chapter presents an overview of relevant terms, theories, and methods. A more in-depth presentation of branched videos and how they differ from linear videos is provided. Terms such as playback bar and stall are explained along with motivations of why it is important to study these aspects. This chapter also presents theoretical aspects relevant for studying digital interfaces, how humans interact with them, how certain events change the user expe-rience, and how the study of these aspects may be conducted. Initially, insights of how an interface can and should support the user are presented, especially exploring the concept of cognitive artefacts. This is followed by research of video stalls, and how they have shown to affect user behaviour. Lastly, a review of previous work about mouse behaviour is given, as well as one for mouse-tracking as a method to study user behaviour.
2.1
Supporting the User
A general purpose of studying an interface and its components is to be able to identify prob-lems in order to solve them and make the interface more user-friendly. Part of this is to iden-tify how the user interacts with the components, and why they act as they do. A component should have a purpose, it should meet some sort of need that the user has. By supporting the user with well-designed components, the user experience will be better, and the product more user-friendly.
One main aspect of how user-friendly an interface or a component is, is how well it me-diates the information it is supposed to transfer to the user. It is not a matter of whether the system has the information the user needs, but if the system manages to mediate this infor-mation in a way that the user understands. The interface is the layer where the system and the user meet, and it is therefore important that the components are using representations that the user can interpret correctly, or in other words, that the system and the user utilise the same language [5], [6].
The interface of a system can be referred to as a cognitive artefact. The term cognitive arte-fact is used to describe an artificial tool that increases cognitive capabilities of the human using the artefact, by manipulating and visualising information [7]. This way, the artefact has representational qualities, creating an easier understood interface between the user and the original information of the system. Whilst a whole system can be a cognitive artefact, for example a computer, its interface is also a cognitive artefact on its own, and in turn, so are the components within the interface. Different parts of a system may communicate well with each other, but the information exchange with the user is the most important part when ensuring usability.
The world we act in is dynamic, meaning that we always have to observe and analyse the environment, making interpretations of what is happening around us, and how our in-teractions affect the environment. The title of this thesis is taken from the following quote by Gibson [8, p. 223]: "we must perceive in order to move, but we must also move in order to perceive". It references this dynamic process of having to act in order to gain information about the world in which we act. This interplay entails certain expectations in regard to
feed-2.2. Branched Video
forward and feedback between the system and the user. Designing artefacts to aid human cognition is to design components that provide the user with both feedforward and feedback during the processes of executing actions and evaluating them [6]. Feedforward refers to the guidance that an interface or a component should present to the user [7]. Guiding the user is done with affordances in various types of elements in the environment, connecting physical aspects of the element to the user’s expectations [8]. Feedback is the information presented by the system in response to our action. If no feedback is mediated, the system (or the inter-face, or the component) is difficult to understand and use [6]. Hence, visibility of the system’s status is an important aspect of usability [5].
Since the interaction between a user and a system is dynamic, it is important to study dif-ferent situations of this interaction, for example during playback interruptions such as stalls, as in this thesis. The user might need some other piece of information when the situation changes, putting different expectations on what information the interface should mediate [6]. This enlightens the importance of ensuring that the representations actually are communi-cating the information they are designed to do, and the importance of studying the interface during various situations even further.
The interface and its components should complement the user with information about the system that the user otherwise would not have, giving feedback of the system’s current status. A main focus of this thesis is to investigate if a playback bar supports the user during video stalls.
2.2
Branched Video
Branched video, also referred to as nonlinear video [9], homogeneous hypervideo [2], and multi-path video [9], [10], is a type of video that is typically interactive in the sense that the viewer gets to choose between predefined alternatives to influence the plot of the video [3]. The term branched refers to the fact that the video is structured like a tree with at least one branch point where the user may choose between at least two branches. Figure 2.1 displays how a branched video can be structured. This tree structure is what separates branched video from linear video. Linear refers to the more common type of video where the starting point and the finish point, and every segment in between, are constant for every viewing.
The mechanics of branched videos can be implemented in different ways. In hypervideos, as described in [2], hyperlinks are provided in each video, linking it to another video, creating a way to navigate between them. This type of branched video got popular on the streaming service YouTube, where each segment was released as a separate video, linking them together
Figure 2.1: Example structure of a branched video. Grey circles symbolise start and end nodes; white circles symbolise branch points; the red line symbolises that only one path may be chosen.
2.2. Branched Video
Figure 2.2: Display of a branch selection in the video player used in this study. Rabbit hole and Happy bunny are the two alternatives that the user may choose between.
through hyperlinks labelled with the alternatives that the viewer had to choose from. Another option is to take a linear video, divide it into segments, and tag it with timestamps in such a way that the video player can tell which segment to load and present next [3], [10]. This latter type is the one used in this thesis. Of course, this assumes access to a video player that can interpret this specific type of information.
Upon reaching a branch point, the user is exposed to the predefined paths they may choose from. The alternatives are typically presented to the user at a set time before reaching the branch point [3]. The amount of seconds preferred by users has been shown to be around three to five seconds, giving the user enough time to make a decision, without the choices stealing too much attention from the video [4]. However, in the video player used for this project, the choices were presented ten seconds before the branch point, in order to give the user enough time to make a decision. The choices are presented to the user through transpar-ent buttons covering a large portion of the video. An example of such a branch selection is displayed in Figure 2.2. Some existing examples of branched videos are Netflix’s Black Mirror: Bandersnatch and Puss in Book: Trapped in an Epic Tale. While they both stick to binary branch selections, they differ in they way they present the alternatives.
There are different approaches to how the choice is made at a branch point. The following two approaches are described in [10]. A user-driven approach is to present the alternatives to the user who then makes the choice. Another option is a knowledge-driven approach where the player decides for the user, based on the user’s previous choices and other information about the user. The user-driven approach is what makes the video interactive, but it also raises the risk for playback interruptions, since the video will stop if the user fails to make a choice before the video reaches the branch point. Interruptions like these could have a great negative influence on the user’s experience watching the video [3]. Hence, some kind of combination of approaches has been suggested, where the user has the option to choose, but once the video reaches a branch point, the player will make a decision for the user. This decision could either be based on previous knowledge about the user [3] or implementing a default path [4].
In a user study [4], the test was designed so that the video froze at the branch points in order to record the users’ full decision times. However, the participants in the user study claimed to have preferred the player to choose a default path if they had not made a decision before reaching the branch point. Although this was not actually tested in the user study, it was implemented in the system for this project in order to give focus to the stalls occurring
2.3. Video Stall
separate from branch points. The stalls studied in this thesis are not in connection with the branch points.
2.3
Video Stall
A video stall is when the video suddenly stops playing and freezes. Video stalling may be due to network limitations or fluctuations, causing the player to not be able to buffer the video in the speed needed for uninterrupted playback [11]. In branched videos, as described previously in Section 2.2, stalls may also be presented at a branch point if the user has not yet made a decision. However, this is not the type of stall studied in this thesis. A stall is a specific type of situation of watching videos, but still an important situation to consider when designing a video player interface. The interface should present relevant and enough information to support the user in understanding what is happening. These user needs may differ during regular playback and stalls.
Stalls have been shown to have a negative impact on the viewer’s experience of watch-ing the video [11], as well as their user engagement [12], which might result in the viewer abandoning the video completely. Factors such as position, duration, and number of stalls, and how they affect the viewer’s user experience, have been investigated using two differ-ent methods. The first method being Quality of Experience (QoE), both recorded continuously during the video and as an overall rating after the video has ended [11]. The second method was to estimate user experience impairment based on viewers’ self-reported experience [13]. Beginning a video with a stall gave lower ratings of both the continuous and final QoE compared to the ratings for a reference video without initial delay. However, the continuous QoE showed differences between shorter stalls (four seconds) and longer stalls (twelve sec-onds), where the longer delays resulted in lower ratings, but the final QoE were not signifi-cantly different between the two lengths of stall [11]. The impairment of the user experience also indicated a difference in regard to the duration of the initial delay, showing a positive linear relation between the amount of seconds (ranging from two to fifteen seconds) and the effect of the stall [13]. Initial delays longer than two seconds give rise to a viewer behaviour of abandoning the video [14]. Stalls earlier in the video have been reported to be more irri-tating than later stalls, however the impairment of the user experience did not show such a difference [13].
The magnitude of the impact of an initial delay differs in regard to the total duration of the video; for shorter videos (about 1 min) the impact is a lot greater than for longer videos (up to 1 hour), and an increase of the duration of the initial stall has a greater impact on shorter videos [15]. Along these lines, if possible, a viewer tends to abandon the video much sooner if there is an initial delay for a short clip compared to a longer video [14]. Regarding stalls in general, a viewer that has experienced stalls for a time corresponding to one percent or more of the video’s full duration, will on average watch about five percent less of the video than a viewer who has experience no stalls [14]. Overall, the level of tolerance appears to raise with the video’s duration.
Both a higher number and a longer duration of stalls have been shown to impact the user experience of watching a video more negatively, but the relation between these two factors is not straight forward. As shown before, the duration of initial delays is a relevant factor for the stall’s impact on the viewer’s experience. This trend has also been reported for stalls in general, that when the number of stalls were kept the same, the level of impairment got higher when the duration of the stall got longer [13]. The number of stalls, however, has been shown to be significantly more influential of the viewer’s experience than the duration of the stalls [11]. One risk with exposing multiple stalls to each participant, especially when similar in duration and location, is that the participant may learn the pattern and foresee what will happen [11].
2.4. Playback Bar
An additional factor relevant to how much impact a stall has on user experience is the type of content in the video, or more specifically, whether the content is of high or low motion (sport being an example of high motion) [15]. With a fixed number and duration of stalls, the higher the motion of the video is, the bigger the impact of the stall will be. This trend was explained as that of higher expectations on the video quality when watching high motion videos [15].
Lastly, knowledge of the system used to present the video has also been shown to affect the level of tolerance. A better network connection tends to result in a lower level of toler-ance [14], suggesting a higher expectation on the system. The same goes for the opposite, knowledge of a slower network connection leads to a higher tolerance for stall events.
Worth noting is that the studied cases referred to above were conducted with linear videos, so the results might therefore not be transferable to branched videos.
2.4
Playback Bar
The term playback bar refers to the bar at the bottom of the screen showing the progress of the video, how much of the video that is left, and often also how much of the video that has buffered [4]. This component has sometimes been referred to as a time-bar [9] or a timeline [2]. Two different playback bars were used in the experiment presented in this thesis, one for linear videos (see Figure 2.3) and one for branched videos (see Figure 2.4).
The function of a playback bar has been described as "a clear visual way to extract infor-mation about (...) playback progress and buffer levels" [4, p. 1]. This description illuminates the main aspects of why a playback bar is a sure feature in a video player’s interface; it is part of the generic video player that is well-known all over the world. At least, this is true regarding players that present linear videos, but with branched videos, additional aspects must be considered. Since the structure of a branched video is shaped like a tree, visualising it as a straight line might not communicate enough information. This is just one example of how a branched playback bar have additional aspects to consider than that of a linear video. It is differences like this one that makes it important to study a component as the playback bar in branched and linear videos separately.
The playback bar’s function of presenting buffer levels has a clear connection to why this specific component is of interest in this thesis. It is important to understand what function the playback bar has, if any, to the user in the situation of a stall. The playback bar used in this experiment was implemented with the intention of mediating information about playback progress and how much of the video that has been buffered. This was done by using different colours to fill the playback bar: red for what has already been watched, light grey for what is buffered, and dark grey for what is left of the video but has not yet been buffered.
Figure 2.3: Playback bar for linear videos.
2.5. Generic Video Player
2.5
Generic Video Player
Interactive video is a concept that has been around for more than twenty years [16], but it is only recently that the need for a standardised branched video player has been raised [2], [4]. Typically, branched videos have been individually implemented for the platform or the specific video. An argument as to why branched videos has not become mainstream is the lack of a generic video player [4]. In comparison with linear videos, it is more demanding to produce a branched video due to more advance editing and lack of tools to help throughout the process [2], so being able to reuse the same tools and interfaces would ease the developing process.
Part of creating a standardised player is to look at the individual elements within its inter-face. Lindskog et al. [4] created and evaluated a generic playback bar and found that users’ experience of a playback bar in a branched video might differ from how they feel about a playback bar in a linear video. Always knowing where you are and, more importantly, how close to a branch point or the end of the video you are were stated to take excitement away from the experience. In cases where branches are not of the same length, the user might no-tice this in the playback bar and chooses the next branched based on this fact. As mentioned previously, this raises the argument to study the impact of a playback bar in branched and linear videos separately.
Lindskog et al. [4] also found reasons to believe that an additional feature of the playback bar is that it would help the user to identify if a sudden stall in the video is due to low buffer conditions. This was not part of their user study and was therefore not tested, but it raised the question which inspired this thesis.
2.6
Mouse Behaviour
In this thesis, user behaviour is studied through mouse movements. Previous research analysing users’ mouse behaviour has studied many different aspects in various situations. Mouse-tracking as a method has also been evaluated, often in comparison to eye-tracking.
Recording a user’s mouse behaviour through cursor data and mouse clicks has been used to study stress during many different types of task, e.g. decision-making [17] and learning tasks [18]. Mouse-tracking has been suggested to be one of the main methods, together with keyboard-tracking, to study stress in human-computer interaction, when the following mea-surements are used: click accuracy, click duration, number of mouse clicks, mouse move-ment [18]. Other mouse-based measuremove-ments that have been used are frequency of mouse clicks, average length of one move, amount and average duration of pauses, and mouse speed [19].
Studying stress through mouse movements has brought forward possible indicators of stress during computer interaction. Shorter and quicker mouse movements are suggested to indicate stress, following that changing to a more stressful environment has been shown to increase mouse usage [18] or more specifically wrist movements [20]. Multiple mouse-tracking studies has found signs of hesitations, which in turn could be indicators of stress, insecurity, or a higher cognitive load. Hesitation can be manifested through an increased amount of mouse clicks [18], slower and arc-shaped motions [21], and shorter pauses within movements, usually around one-tenth of a second to one second long [22].
Apart from studying stress, mouse-tracking has frequently been used to study user atten-tion, often being compared with eye-tracking. This type of studies have found contrasting results whether mouse-tracking is a good alternative method to eye-tracking regarding esti-mation of gaze fixation. Arguments against this method are that, at least for smaller tasks, it tends to not be enough overlap between gaze fixation and cursor position, and the mouse is mainly used to perform direct actions instead of following the gaze [23]. However, other stud-ies have shown a greater gaze-cursor alignment, but only during what is called active mouse-use. When the user actively uses the mouse, there is an average distance between gaze
fixa-2.6. Mouse Behaviour
tion and cursor position of 130 to 280 pixels [24], which could point to that cursor position is a good estimation of where the user’s attention is directed in general. However, even if over-lap is recorded, the gaze tends to vary much more than the cursor, although often around the cursor’s position, placing it to some extent at the centre of attention [25]. Inactive mouse-use refers to the time when the user explores the screen without the cursor present or when the user takes time to read or reflect over future actions, often recorded in data as longer time periods where mouse speed is zero. During this inactivity, the cursor is not thought to be a good indicator of where the user’s attention is directed [24].
By using comparisons between gaze fixation and cursor position, it has been suggested that both gaze and cursor movements are affected in similar ways. While reading with dis-tracting elements next to the text, a reader moved both gaze and cursor with back-tracking motions [26], hence showing similar behavioural patterns for both aspects, which suggests that reading with distracting elements changes the behaviour to more actively using the mouse. The back-tracking motions were also suggested to be signs of frustration [26]. In contrast to this back-tracking pattern and the hesitation patterns before, cursor movements that are straighter and going directly to an area of interest has been argued to show intention-ality and familiarity with the task [27].
Mouse-tracking has also been used to study emotion in human-computer interaction. Similar to the results described above in relation to stress and hesitation, negative emotions have been found to impact mouse behaviour more than positive emotions, compared to a neutral state [25]. Along the same lines, user’s mouse behavioural pattern differs when they listen to happy and sad music compared to neutral, although only for men [28]. The male users’ mouse patterns changed during the test depending on the type of music that had been played, but the overall performance of the task had not changed. In general, the women tended not to move the cursor in a straight line to the target, whereas the men did, and the women also had a tendency to backtrack more often than the men [28]. Their suggestions for this gender difference were based on differences in spatial-motor ability, differences in which cognitive strategy to use, and the fact that male behaviour, in general, tend to be more risk-taking [28].
Based on Attentional Control Theory (ACT), Hibbeln et al. [29] set out to explain why emo-tions has the effect on mouse movements as it has previously been shown to have. According to ACT, anxiety can affect the user’s ability to control their attention [30], due to distracting elements competing for attention. In turn, attention has been found to affect movement [31], and therefore also mouse movement. In [29], they hypothesised that negative emotions lead to reduced ability to execute precise movements, resulting in longer distances moved. They also hypothesised that negative emotions would require more mental resources, based on [30], resulting in lower speed of the mouse movements. Both these aspects were shown to be significantly different, along the lines of the hypotheses, between the baseline condition and the condition with induced negative emotions. In addition to this, emotions were sig-nificantly deduced from mouse movements [29], suggesting that mouse-tracking may be a justified method for estimating users’ emotions. From ACT, it also follows that the user’s be-haviour switches from goal-directed to stimulus-driven due to negative emotions [30], mean-ing that the user’s attention is attracted by the presence of different elements, rather than controlled by the user to fulfil a specific goal.
Mouse-tracking can be used in combination with other methods, for example adding a quantitative dimension to an otherwise qualitative-focused user study, since mouse be-haviour can be recorded without the user being aware of it [29].
3
Method
In this chapter, the design of the experiment, the setup, the participants, and the analysis are described. The data collection for this thesis was conducted together with another project, which is why some parts of the procedure include aspects not covered in this thesis.
3.1
Experimental Design
This experiment aimed to study mouse movements through two different dimensions regard-ing watchregard-ing videos. The first dimension was whether the video player had a playback bar or not; the second dimension being the type of video, either branched or linear. By merging these two dimensions, four different combinations emerged, as presented in Figure 3.1. To aid comprehension, the following abbreviations will be used to easily distinguish between videos of different types: BP refers to a branched video with a playback bar, BN refers to a branched video without playback bar, LP refers to a linear video with a playback bar, and LN refers to a linear video without a playback bar. These abbreviations will also be used com-bined, for example BPLP refers to the group that first watched two branched videos with a playback bar, and then two linear videos with a playback bar.
From these four combinations, three aspects were brought forward as most important for answering the research questions of this thesis. The three aspects were the following: differences between linear and branched videos with a playback bar present, differences be-tween having a playback bar or not during branched videos, and differences bebe-tween having a playback bar or not during linear videos. From these three aspects, six groups were
3.2. Participants
Figure 3.2: The six groups, based on three aspects, that the participants were divided into. Playback bar is abbreviated to pbb.
ated, approaching each aspect from opposite directions. For example, one group, providing data for the first aspect, starts of by watching branched videos with a playback bar (BP) fol-lowed by linear videos with a playback bar (LP), whilst a second group starts with the linear videos (LP), finishing with the branched ones (BP). See Figure 3.2 for a representation of all six groups, using the abbreviations from above. By alternating the order of the videos, it is possible to test each aspect for order effects.
3.2
Participants
Data were collected from a total of forty-six participants. However, due to various technical issues, thirteen of these participants had to be removed from the data set. Some data from the remaining participants were also problematic to use, and therefore the exact number of participants included when conducting statistical tests will be presented in connection to each specific test in Chapter 4. In addition to this, one participant decided to withdraw from the study.
The data used for the analysis came from a total of thirty-two participants (fifteen women, seventeen men) in the ages of 21-28 (M = 23.75, SD = 1.70). Most participants were univer-sity students, although from varying fields of education, where Information and Commu-nication Technologies (twelve participants), Engineering, Manufacturing and Construction (seven participants), and Cognitive Science (six participants) were the most frequent. All participants had good knowledge of both Swedish and English. Recruitment of participants was conducted through convenience sampling. To ensure equal sample sizes within each group whilst still, to some extent, randomly assign a group to each participant, a round-robin approach was used.
The participants had varying habits of how many hours spent, on average, watching videos and playing computer games on a weekly basis (see Figure 3.3). The participants were asked to exclude watching TV and going to the movie theatre, since, in these situations, the act of watching a video is more passive than it may be in other situations. Playing computer games is also a more active process of interacting with stories.
For the estimation of hours spent watching videos, the most common answer was "3-7h" with ten participants, and then "8-15h" with eight participants. No participants answered "Never", and one participant answered "31-35h". There were participants with varying habits of watching movies in each group. The three participants who claimed to watch 26+ hours of videos each week belonged to three different groups belonging to each of the three as-pects studied. The twenty-three participants who answered fifteen hours or less were evenly divided between all six groups.
3.3. Equipment
Figure 3.3: The participants’ weekly habits of watching videos and playing computer games.
The participants’ weekly habits of playing computer games differed from their habits of watching videos. Thirteen participants answered "Never", making it the most common an-swer. The median answer was "0-2h", and one participant gave the answer "31-35h". The thirteen participants who reported to never play computer games on a weekly basis were evenly spread out between all six groups, and the two participants who reported 26+ hours belonged to two separate groups and aspects. The rest of the participants were spread out between all six groups, although with a majority in the groups containing branched videos.
In addition to this, the participants also had varying previous experience regarding branched videos. Fifteen participants claimed to have never seen a branched video before tak-ing part in this experiment. Another fifteen reported to have watched one to three branched videos beforehand. The additional two participants answered that they had watched three to five branched videos previously. These three levels of previous experience with branched videos were evenly divided between all six groups. Six participants reported to have seen Black Mirror: Bandersnatch previous to the experiment, another participant mentioned Net-flix in general, and two participants referenced hypervideos on YouTube. One participant mentioned having seen branched videos on both YouTube and Netflix.
3.3
Equipment
The equipment used in this experiment constitutes of a laptop, a mouse, and a video player created specifically to present branched videos. Digital online questionnaires were used to gather demographic information about the participants and their self-reported experi-ence during the experiment. Mouse-tracking was conducted automatically through the web browser during the videos.
3.3.1
Setup
The computer used during the experiment was a laptop (Asus X550J) with a 15.6-inch, non-touch, 1920x1080 screen, an Intel processor (i7-4720HQ), that was running Windows 8.1. The mouse used was a wireless optical mouse (38-5512/SM-356AG), and the computer setting for mouse speed was set to medium speed on a three point scale, with no acceleration, during the entirety of the experiment. The laptop’s touch pad was deactivated during the whole experiment.
The experiment was conducted through a web browser, with one tab for each question-naire, video, and instruction. An online questionnaire service was used to collect information
3.3. Equipment
from the participants. The videos were shown in as close to full screen as possible (150 percent zoom in the web browser). The video player will be described in more detail in Section 3.3.2. All information was given to the participant via HTML5 web pages.
Two keys on the keyboard were used as controls during the experiment. The right arrow on the numpad was used to proceed to the next step in the experiment (i.e. switch to the next tab in the browser). This action was deactivated during the videos so that the participant would not be able to skip a video. Instead, a new tab was automatically opened when the video ended with information about the next step. The key "P" was used to start the video once it was shown on the screen. Both keys were highlighted with bright orange washi tape and marked with a black right arrow and a black "P", respectively.
3.3.2
Video Player
The video player used in this experiment was made public with the publication Generalized Playback Bar for Interactive Branched Video from 2019 [4]. Their system is based on a stream-ing technique called Dynamic Adaptive Streamstream-ing over HTTP, or DASH, which through dash.js is implemented in JavaScript language, enabling connection with HTML elements. They adapted the video player to be able to present branched videos, also adding a playback bar showing the progress and the buffer level of the video.
The playback bar is shaped based on the structure of the current video, showing branches if there are any (see Figure 2.3 and Figure 2.4). With the use of colours the playback bar offers information about how much of the video that the user has seen (red), how much of the video that is buffered (light grey) and what is left of the video but not yet buffered (dark grey). Throughout the experiment the playback bar was used in two different ways. During BN and LN videos, the playback bar was inactivated completely. During BP and LP videos, the playback bar was present when the cursor was moving and, if it was a branched video, when the choices before a branch point were shown on the screen.
3.3.3
Videos and Stalls
The eight different videos that were presented to the participants were cut and combined from the open source linear video Big Buck Bunny [32] (with Creative Commons Attribution 3.0 license). The decision to use the same original video for all the videos used in the exper-iment was based on the findings that stalls may affect user experience differently between videos with higher and lower level of motion [15]. The video player reads a meta file stating the starting point and the duration of each video segment, as well as how the segments are combined at branch points. In this experiment, each branch point was binary, with only two branches to choose between. Four branched videos and four linear videos were prepared. The linear videos were encoded as branched videos with only one segment, and with no branch points.
Meta files were also created to implement stalls during the videos, ensuring that the posi-tion and the duraposi-tion of the stalls were consistent between condiposi-tions and participants. The video player is also implemented so that the buffer in the playback bar stops at the position of the stall, only showing again once the stall is over.
Each linear video and its correspondent branched video were composed to include the same part of the original video, giving the participants the same experience between condi-tions. However, the branched videos had to include more footage since the two branches at each branch point were supposed to be giving the participant different plots.
The first video (see Figure 3.4) was 160 seconds long, contained two stalls, both 18 seconds long, placed at 64 seconds and 128 seconds, resulting in a total duration of 196 seconds (or three minutes and 16 seconds). The branched version of the first video had branch points at 32 seconds and 96 seconds, presenting one branch point before the first stall. The second to fourth video (see Figure 3.5) were all of the same structure, with stalls at the same position,
3.3. Equipment
Figure 3.4: Graphic representation of the structure of video 1
Figure 3.5: Graphic representation of the structure of video 2-4
with the same duration. This decision was based on the findings that changes of these factors may influence the user experience, creating different stall situations [11], [13]. These videos were 72 seconds long, with a 24 seconds stall after 48 seconds. The branched versions had one branch point at 40 seconds. The total duration of each video was 96 seconds.
3.3.4
Mouse-tracking
During all videos, mouse-tracking was used to gather data about the participants’ mouse behaviour. Cursor movements were recorded through the web browser during the whole video, saving all data in a csv file once each video ended. The measurements saved were the following:
• time, in milliseconds, from the start of the video (when the participant pressed "P"), with data points roughly every 16 ms,
• x and y coordinates of the cursor,
• speed, in pixels per second, calculated by comparing x and y coordinated between two subsequent data points, and
• stalls, recorded as true or false, based on whether a stall was currently happening or not.
Based on these recorded measurements the following variables were calculated, preparing for the analysis:
• percentage of mouse activity during stalls, based on how many of the data points registered during a stall had a speed value that was separate from 0,
• percentage of mouse activity during branch selection, based on how many of the data points registered during the ten seconds before a branch point had a speed value that was separate from 0,
• percentage of mouse activity during regular playback, based on how many of the data points registered during a regular playback had a speed value that was separate from 0, • average speed during active mouse-use during stalls, that is, not including data points
with a speed value of 0,
• average distance from the playback bar during stalls, comparing the y coordinate of the cursor with the y coordinate for the bottom of the playback bar, and
• total distance moved during stalls, based on calculating the hypotenuse of the x and y coordinates between two subsequent data points.
3.4. Procedure
3.4
Procedure
The experiment was conducted in a total of six steps with four main steps, each containing one video and one to three questionnaires. These four steps were preceded by an introductory step containing information about the experiment in general, instructions for the participant, and a questionnaire collecting demographic information about the participant. The experi-ment ended with a final step where the participant got debriefed about the true purpose of the experiment. The whole experiment took about 30 minutes, although some participants only spent around 20 minutes, whilst others took up to 50 minutes.
In the introductory step, the participant received information about the experiment in general and specific instructions for what they were to do during the experiment. It was explained that the purpose of the experiment is to study user experience of two different types of videos. Both linear and branched video were explained to the participant, although they were told that it was not sure that they would be watching both types of video. The participant was told that the videos would not include any audio, and that while the ques-tionnaires were written in English, both Swedish and English were acceptable for answers. A consent form was presented and explained to the participant, the participant was able to ask questions, and once the consent form was signed the experiment started. The controls were introduced to the participant, explaining the "P" key and the right arrow on the numpad. The test supervisor left the room and the participant started filling out a questionnaire collecting demographic information including gender, age, level and field of study, average number of hours spent weekly watching videos (excluding TV and cinema) and playing computer games.
The first main step started with a Short Stress State Questionnaire (SSSQ), which is not included in this thesis. Then the participant received the information that they were to watch a video and would be asked questions about the video at the end of the experiment. This was a false task created to encourage the participant to be active during the video. The first video was shown, followed by a NASA Task Load index (NASA-TLX) questionnaire, which is also not included in this thesis.
In the second step, the information given to the participant was that a problem would happen and that they were to try to understand what that problem was; the problem referred to, of course, being the stall. The video was shown and was once again followed by a NASA-TLX questionnaire, and then another SSSQ. The next step was a questionnaire with open-ended questions about what the problem was, what they wanted to do in order to find out what the problem was, which steps they actually took, and if some additional information could have helped to understand the problem more easily.
The task in the third step was the same as in the second, only stating that the video player would be different, but not in what way. This difference was either a change between linear and branched videos, or with or without playback bar, but the participant was unaware of the difference beforehand. After the video, the participant filled in another NASA-TLX ques-tionnaire, as well as the same questionnaire with open-ended questions as in the previous step.
In the fourth step, the participant was given the information that a stall would occur and that their task was to identify why the stall happened. In this step, the video was followed solely by the questionnaire with open-ended questions from the two previous steps.
The final step included a short questionnaire about the participant’s previous experience with branched videos, how many they had seen before this experiment, and, if any, the name of them. The participant was then asked to summon the test supervisor, who gave a de-briefing, once again explaining the aim of the experiment. This time also mentioning the mouse-tracking and asking for permission to use this data as well. The participant had the opportunity to ask any question about the experiment.
3.5. Pilot Studies
3.5
Pilot Studies
Two pilot studies were conducted, the first with four participants (21-26 years, M = 23, SD = 2.2; two women, two men), and the second with three participants (23-27 years, M = 25, SD = 2; three men). All participants completed the whole experiment from start to finish, which made it possible to time the duration of the experiment. Afterwards, the participants were asked some additional questions about the experiment, ensuring that instructions and questionnaires were understandable. The participants were divided into different groups making sure that all videos worked properly.
After these pilot studies, the following changes were made. Information about the two different types of video included in the experiment was given in the introductory phase in order to prepare the participant for what they might experience. This was thought to raise the probability for natural behaviour if they encountered branched video for the first time.
Originally, the tasks of Step 2-4 were combined and conducted through two steps, the task being that the participant were to identify the problem as well as the reason for it. These were then divided into the separate steps presented in Section 3.4. This way, the procedure was structured in more of a step-by-step way, giving the participant one new aspect of the task in each step.
The duration of the videos as well as the number and duration of the stalls were also al-tered. In the first video, three shorter stalls were changed into two longer once. This decision was based on the participant not having enough time to react to a stall, hence making them longer, and reducing the number of stalls in order to not make the video unnecessary long. The other videos where made longer so that the videos would not end immediately after the stall.
Changes were also made to how the mouse-tracking was documented, making sure that the files showed the true duration of each video in milliseconds. Finally, the key "P" was deactivated, making it impossible to pause the video once it was started.
3.6
Analysis
During the experiment, both quantitative and qualitative data were gathered. The quantita-tive data were analysed through each of the three aspects separately (i.e. BPBN and BNBP, BPLP and LPBP, and LPLN and LNLP). Wilcoxon signed-rank test were conducted on these groups to study the variance. Mann-Whitney U tests were used to compare user behaviour between groups, where each participant only belonged to one group. Non-parametric tests were used due to the small sample sizes and the fact that the samples varied in sizes. The small sample sizes also led to the usage of exact p-values instead of asymptotic.
The qualitative data were analysed through thematic analysis, however in a restricted manner. Each participant’s answers were coded and grouped together into themes and sub-themes. Answers not regarding the video player’s interface or the user’s interaction with it were categorised together since the purpose of the analysis was directly connected to how the participants interacted with the video player’s interface. Afterwards, the themes were analysed to identify if some were more common within certain groups compared to others. Since the playback bar is of high interest in this thesis, aspects regarding this component were highlighted and focused on.
4
Results
This chapter is divided into subsections presenting results of mouse behaviour during stalls from the three studied aspects: having a playback bar or not, both during branched and linear videos, and differences between branched and linear videos. Then the focus shifts to the whole video, to compare user behaviour during the stall to the other parts of the video. Finally ending with a thematic analysis of the participants’ self-reported mouse behaviour.
For the statistical tests, the data were divided into three combined groups: the two groups of participants watching branched videos with a playback bar as their first two videos fol-lowed by two branched videos without a playback bar (BPBN), and vice versa (BNBP); the two groups watching linear videos with and without a playback bar (LPLN and LNLP); and finally, the two groups of participants that watched two branched and two linear videos, al-ways with a playback bar present (BPLP and LPBP). Conducting the analyses in this way focuses each comparison on one feature at a time: if there is a difference between having a playback bar or not, both during branched and linear videos, and if there is a difference between the two types of videos with the playback bar being the fixed variable.
Within these groups, the following main dependent variables were compared and anal-ysed. The participant’s mouse activity during the stalls and how that compares to the rest of the video, were used to give an overall sense of how much movement the participant en-gaged in. Average cursor speed, average distance from the bottom of the playback bar (or, in the cases where no playback bar was shown, the location where it would have been), and the total distance that the cursor moved on the screen, were all analysed during each stall.
The variables presented above were studied through each participant’s second and third video, giving data from alternating videos (e.g. video 2 being branched with a playback bar and video 3 being linear with a playback bar) whilst keeping the task fixed. Due to the technical problems during the data collection, the sample sizes varied between the tests. The size of each specific sample is reported next to the corresponding test. In all cases, the exact p value was calculated and used due to small sample sizes, and a 95 percent confidence interval was used. Abbreviations for the videos will be used, see Section 3.1 for an explanation of these.
4.1
Difference Between Having a Playback Bar or Not
This section addresses the aspect of how a playback bar may influence mouse behaviour during stalls. This is initially approached from the perspective of branched videos, followed by linear videos. Finally, the groups were combined to study if there were any overall order effects of having a playback bar before or after not having one.
4.1.1
Playback Bar During Branched Videos
Starting with the group of participants that only watched branched videos, either with a playback bar present during the first two videos (BPBN, N = 6) or the last two videos (BNBP, N = 6). When combining all BP videos and all BN videos, no matter if it was the participant’s
4.1. Difference Between Having a Playback Bar or Not
second or third video, there was no significant difference between how active each participant were during the stalls: BP (Mdn = 0.017) and BN (Mdn = 0.023), z = -0.187, p = 0.910, r = -0.027. Three participants in this combined group did not move the cursor at all during the stalls, in neither video 2 nor video 3. An additional three participants showed full inactivity during the stalls of one of the videos, two being BN and one BP. From the wider perspective of all videos within this combined group, six out of twenty-two of all BP videos (27.3 percent) presented full inactivity, and the same thing was observed for BN videos, six out of twenty-two (27.3 percent).
The analysis of this group continues without the three participants who were inactive in both their videos, which updates the sample sizes of group BPBN (N = 5) and group BNBP (N = 4). For the active participants, there was no significant difference of their av-erage cursor speed during active mouse usage: BP (Mdn = 424.698) and BN (Mdn = 515.847), z = -0.415, p = 0.734, r = -0.138. Similarly, there was not any significant result regarding total distance moved on the screen during stalls between the groups BP (Mdn = 695.282) and BN (Mdn = 689.517), z = -0.415, p = 0.734, r = -0.138.
Finally, a look at the average distance between cursor and the playback bar (or where the playback bar should have been) during stalls. Due to inactivity during one of the videos, the sample sizes need to be adjusted further (BPBN, N = 3; BNBP, N = 3). There was, however, no significant difference for this variable between BP (Mdn = 237.473) and BN (Mdn = 201.267), z = -0.943, p = 0.438, r = -0.272.
4.1.2
Playback Bar During Linear Videos
The combined group of participants watching linear videos with or without a playback bar contained of LPLN (N = 2) and LNLP (N = 4). Considering the second and third video, no significant difference was observed of the participants’ mouse activity during the stalls: LP (Mdn = 0.000) and LN (Mdn = 0.000), z = -1.342, p = 0.500, r = -0.387.
Four out of the six participants in this combined group, that only watched linear videos, never moved the cursor during the stalls. The other two participants only moved the cursor during the stalls of their third video, which in both cases were LP. Hence, in regard to the stalls of video 2 and video 3, complete inactivity was observed in all LN videos. For the LP videos, four out of six videos showed no mouse activity at all during the stalls.
Taking all four videos into account, eleven out of twelve LN videos (91.7 percent) showed no mouse activity at all during the stalls. Regarding LP videos, on the other hand, seven out of twelve videos (58.3 percent) showed no active mouse movement at all during stalls. Due to this high level of inactivity, average speed, average distance to the playback bar, and the total distance moved could not be analysed for these groups.
4.1.3
Order Effects of Switching Between Conditions
The four groups BPBN, BNBP, LPLN, and LNLP were studied in order to analyse if there were any order effects in general when switching between having a playback bar or not, in both directions. The groups BPBN and LPLN were combined, as well as BNBP and LNLP, creating two new groups: XPXN and XNXP. Then the aspects of having a playback bar or not were studied separately, comparing mouse movements of videos with a playback bar (here called XP videos) between the participants who watched their XP videos as their first two videos and those who watched them as their last two videos. The same goes for videos without a playback bar, here referenced as XN videos. All four of the participants’ videos were used, even if the tasks are different for the four videos, apart from video 2 and video 3. However, the larger sample is evenly balanced between groups going from having a playback bar to not having one, and the group going in the opposite direction.
The first aspect analysed was the aspect of having a playback bar (XP). Studying mouse activity during stalls between the two groups XPXN (N = 16) and XNXP (N = 18) showed no
4.2. Difference Between Branched and Linear Videos
significant difference between having XP as the first two videos (Mdn = 0.006) and having XP as the last two videos (Mdn = 0.046), U = 105.000, z = -1.385, p = 0.174, r = -0.237. Out of these participants, seven from the group XPXN and six from the group XNXP had an activity of zero percent.
With the updated sample sizes, the analysis continues to only look at the active partic-ipants in group XPXN (N = 9) and group XNXP (N = 12), from the perspective of average speed, average distance between cursor and playback bar, and total distance that the cursor moved on the screen.
There was a significant order effect of average speed during stalls between active par-ticipants starting watching XP videos (Mdn = 870.180) and those who watch them last (Mdn = 394.191), U = 26.000, z = -1.990, p = 0.049, r = -0.434. The same goes for average distance between cursor and playback bar during stalls for the active participants watching XP for the first two videos (Mdn = 264.975) and those who watched XP for the last two videos (Mdn = 140.384), U = 17.000, z = -2.629, p = 0.007, r = -0.574. Finally, there was no significant difference in total distance moved between having XP as the first two videos (Mdn = 66.748) and having XP as the last two videos (Mdn = 497.048), U = 105.000, z = -1.385, p = 0.174, r = -0.237. The results show that the user moves the cursor slower and closer to the play-back bar when watching videos with a playplay-back bar are presented after videos without one (XNXP) compared to watching videos with a playback bar first (XPXN).
The second aspect studied was the aspect of not having a playback bar present (XN). Beginning again with mouse activity during stalls between the same groups as before: XNXP (N = 18) and XPXN (N = 16). This time, there is a significant order effect of mouse activity during stalls between watching XN videos as the first two (Mdn = 0.000) and the last two (Mdn = 0.038), U = 84.000, z = -2.213, p = 0.027, r = -0.380. Of these participants, eleven from group XNXP and six from group XPXN showed hundred percent inactivity. This meant that the samples sizes needed to be updated before further analysis: group XNXP (N = 7), group XPXN (N = 10).
Focusing on the active participants, the following results were observed. No significant order effect was observed for average speed during stalls between the participants watching XN videos first (Mdn = 667.539) and those who watched XN videos last (Mdn = 947.527), U = 26.000, z = -0.878, p = 0.417, r = -0.213. Regarding average distance between cursor and where the playback bar would have been, there was, however, an order effect (XN first, Mdn = 165.647; XN last, Mdn = 249.726), U = 12.000, z = -2.245, p = 0.025, r = -0.544. The same goes for total distance moved on the screen (XN first, Mdn = 433.442; XN last, Mdn = 1682.333), U = 14.000, z = -2.049, p = 0.043, r = -0.497. These results indicate that the user is more active, making longer moves further away from the playback bar when watching videos without a playback bar after watching videos with one (XPXN) compared to watching videos without a playback bar first (XNXP).
4.2
Difference Between Branched and Linear Videos
This section analyses the aspect of switching between branched and linear videos, using the two groups BPLP (N = 6) and LPBP (N = 3). Within this group, studying only video 2 and video 3, there was no significant difference regarding mouse activity during stalls between BP (Mdn = 0.019) and LP (Mdn = 0.019), z = -1.014, p = 0.375, r = -0.239.
Looking at video 2 and video 3 for all participants in this combined group, two partic-ipants showed inactivity during the stalls of both videos. An additional participant never moved the mouse during the stall of their first video, being a BP video. Including all four of the participant’s videos, the number of BP videos with complete inactivity was five out of sixteen (31.3 percent), and three out of sixteen for LP videos (18.8 percent).
With these updated sample sizes, BPLP (N = 5) and LPBP (N = 2), average speed and total distance moved were analysed. No significant difference of average cursor speed was
