Human Performance Modeling
- 1 Slides
- 2 Readings
- 3 Reading Critiques
- 3.1 Vineet Raghu 14:09:28 9/17/2015
- 3.2 Adriano Maron 18:40:36 9/19/2015
- 3.3 Xinyue Huang 0:32:23 9/20/2015
- 3.4 Ameya Daphalapurkar 10:22:39 9/20/2015
- 3.5 Manali Shimpi 11:16:53 9/20/2015
- 3.6 Matthew Barren 16:04:57 9/20/2015
- 3.7 Kent W. Nixon 17:47:53 9/20/2015
- 3.8 Mingda Zhang 17:58:58 9/20/2015
- 3.9 Chi Zhang 22:58:06 9/20/2015
- 3.10 Priyanka Walke 0:08:46 9/21/2015
- 3.11 Shijia Liu 0:18:29 9/21/2015
- 3.12 Zihao Zhao 0:24:04 9/21/2015
- 3.13 Lei Zhao 1:10:58 9/21/2015
- 3.14 Samanvoy Panati 1:18:19 9/21/2015
- 3.15 Zinan Zhang 3:43:22 9/21/2015
- 3.16 Darshan Balakrishna Shetty 7:09:19 9/21/2015
- 3.17 Ankita Mohapatra 8:03:06 9/21/2015
- 3.18 Jesse Davis 8:27:39 9/21/2015
- 3.19 Sudeepthi Manukonda 8:48:46 9/21/2015
- Beating Fitts' law: Virtual enhancements for pointing facilitation, Ravin Balakrishnan, International Journal of Human-Computer Studies, 61(6). p. 857-874.
- Beyond Fitts' Law: Models for Trajectory-Based HCI Tasks, Johnny Accot, Shumin Zhai Proceedings of CHI 1997, p295-302
- An error model for pointing based on Fitts' law, Wobbrock, J.O., Cutrell, E., Harada, S. and MacKenzie, I.S. Proceedings of CHI 2008, pp. 1613-1622.
Vineet Raghu 14:09:28 9/17/2015
Beating Fitts’ law: Virtual Enhancements for Pointing Facilitation This paper is a survey of studies from 2004 and earlier that relate to ways to improve pointing task times in interface design. In particular, the paper discusses how standard pointing techniques are limited by Fitts’ Law, which describes that pointing time is limited by the width of the target and the distance to the target. Thus, the paper details two methods to improve pointing time, which are to increase the width of the target or to decrease the distance to the target. The survey of techniques that were modern at the time of writing is very thorough and clear. The paper also gives sufficient background as to how Fitts’ Law could be overcome in theory. However, I’m unsure as to the current significance of the work described here. Most of these techniques to improve pointing times do not appear to have been implemented into standard operating systems, and moreover, many current interfaces have varied input devices that do not involve a virtual cursor e.g. phones, tablets, etc. For these touchscreens, it is difficult to determine the movement of the finger while it is off the screen to implement many of the design techniques described in the paper including increasing the size of neighborhood icons or increasing C-D gain based upon trajectory prediction. Also, it is unclear to me in what domains these performance improvements in pointing would make a significant difference to users or if these are simply a theoretical concern to HCI researchers. I would imagine the performance gains for many of these solutions are on the order of milliseconds which would not be noticeable to the standard interface user, and it would be a nice addition for the paper to discuss some specific results from the studies so readers would know the improvements achieved by these techniques. ----------------------------------------------------------------------------------------------------------------------------- Beyond Fitts’ Law: Models for Trajectory – Based HCI Tasks The authors of this paper present a general mathematical model for predicting movement time in trajectory based tasks such as steering. This model is meant to provide a basis to evaluate more modern input devices that accomplish more complicated tasks than the simple pointing of Fitts’ Law. I found it very interesting though that the final model that they derive is a pretty natural extension of Fitts’ Law to the steering experimental setup, since it is a linear function of the index of difficulty, and the index of difficulty is still defined based upon the amplitude and width of the task. This result appears to be a very important result in the context of HCI research, since it provides a link between the past input technology and modern input technology. Using this framework, HCI researchers can now compare modern input devices that accomplish more complicated tasks than their predecessors. I did not find any particular flaw in this paper, though I do believe that some more mathematical explanation could be useful for readers, since there are times where the paper leaves out some non obvious derivation.
Adriano Maron 18:40:36 9/19/2015
Beating Fitts' law: Virtual enhancements for pointing facilitation: This paper advocates that the Fitts' law, which in essence calculates the effort (movement time - MT) to acquire a target in the real world, does not impose hard constraints when selecting targets in a virtual interface. Therefore, it becomes possible to reduce the MT by changing the distance to targets (D), or change their size (W). The authors survey techniques that: (1) decrease D; (2) increase W; (3) both decrease D and increase W. Fitts' original work considered that physical targets were selected by direct indication using the human hand. For the studies in the virtual environment, the assumption in that pointing involves an intermediary controlling device, such as mouse or touchpad. Such assumption was completely acceptable in 2004. However, nowadays there is the touch/multi-touch interaction option, which raises a whole new set of variables, such as: precision, hidden targets, physical distance to targets, etc. Studies considering those variables are essential, given that most of the current devices provide such interaction option. Some of the surveyed techniques for reducing D include a redesign of menus and widgets, where items could be arranged in a pie format, instead of a list format. Such change may be applicable in certain contexts (pop-up menu, grouping icons), but it is certainly not a general approach for every scenario. Object Pointing, which is one of the techniques for removing the empty space between the cursor and the target, results in a unnatural behavior where the cursor moves differently from the mouse/touchpad, violating our conceptual model that the cursor behaves almost exactly as the control mechanism. Another way of facilitating pointing is to increase W. Area cursors are a good approach, as long as there are clear visual clues about changing in the cursor's behavior. One of the most successful techniques for increasing W is based on Expanding Targets, and its Fish-Eye implementation. Expanding targets and moving their position on screen is a good alternative for small displays where all the information can not fit in its original (selectable) size. In this scenario, targets selected more frequently have bigger W and a more central position on screen. The last approach for facilitating pointing tries to decrease D while increasing W. Dynamic control-display gain, also known as "mouse acceleration", changes the rate between the speed of the cursor with respect to the speed of the controlling device. This has been successfully used in many user interfaces, and is capable of improving the pointing performance. In conclusion, this paper provides an easy to read survey of techniques used 10-15 years ago, that nowadays can be extended to consider the new interaction options we have available. ======================================== Beyond Fitts' Law: Models for Trajectory-Based HCI Tasks: This paper proposes a mathematical model to describe the "steering law", in which is possible to calculate the time to perform a trajectory (e.g., hierarchical menu navigation) based on the length of the trajectory (how many items in the menu) and its constraints (width of the menu, different levels, ...). This model provides means for a direct comparison between different UI features, such as list menus vs pie menus. Based on the Fitts law' claim that the faster we move, the less precise our movements are, and in the inability of the Fitts' law to model trajectory-based tasks, the authors conducted a series of experiments and derived a model to predict the movement time (MT) such tasks. The experiments followed the "steering between boundaries" paradigm, and asked subjects to use a stylus pen to perform trajectory-based tasks with increasing level of difficulty. The model identified that there is a linear relationship between the MT and the "tunnel" width in steering tasks. More importantly, this model helps in the evaluation of of user interfaces in terms of difficulty and length of the movements. One caveat of the experiments is that they were entirely based on a stylus pen controller. Using mouses or touchpads can lead to less accurate trajectories, which could influence the results and, consequently, the model definition. It would be important to execute the experiments with all 3 different control methods and compare their results to fully validate the model.
Xinyue Huang 0:32:23 9/20/2015
“Beating” Fitts’ law: virtual enhancements for pointing facilitation The paper introduced different technologies for artificially enhancing pointing performance in graphics user interface. It analyzed completely different technologies and compared them with Fitts’ law to elaborate the advantages and disadvantages of these technologies. Fitts’ law asserts that the movement time of pointing is related with the width of target and the distance. So we can improve the pointing performance from three aspects: 1) attempt to decrease D, 2) attempt to increase W and 3) both decrease D and increase W. To decrease D, there are some technologies such as designing widgets that minimizes D or temporarily bringing potential targets to the cursor (drag-and-pop). There exists some disadvantages, for example, drag-and-pop would annoy the user or even interfere with the user’s ability to select an existing nearby target. Another technology to decrease D would be removing empty space between the cursor and targets. To avoid undesirably removing all the empty space. A technology called object pointing is designed. Another aspect is to increase W. Related technologies would be developing area cursors rather than pointing cursors. It aimed to alternatively increase the width of cursors instead to lower index of difficulty for targets. Another technology is to expand targets, which has been applied in modern products such as MacOSX “dock”. The final aspect is to both decrease D and increase W. Design efficiencies could be obtained by dynamically varying the C-D gain based on input device speed, which is the ratio of the amount of movement of an input device and the controlled objects. Though C-D gain adaptation can improve pointing performance to a certain degree, especially for isolated targets, it does not work well for multiple target presented. There are a lot of technologies developed and some of them are promising and are efficient compared with Fitts’ law. However, none of them can work well in all types of situations encountered in graphical user interface. The paper also mentioned that user acceptability is more important than quantitative performance in human computer interaction. Beyond Fitts’ law: Modeling for Trajectory-Based HCI Tasks The paper found that Fitts’ law cannot successfully model trajectory-based interactions and then explored a simple regularity in trajectory-based task called “steering law”. The paper first stated that Fitts’ law is readily appreciated by HCI but it is not an adequate model for trajectory-based tasks. Then considering the success of Fitts’ law, the paper tried to construct a regularity to quantize the trajectory-based task. The paper then designed four kinds of experiment. The first one is goal passing and set two goals and the result shows that the steering task follows the same logarithmic law as Fitts’. Then they designed the second experiment and extended the number of goals to infinity and transfer the problem to “tunnel traveling”. Then paper then defined the third experiment and added the width constraint of the tunnel. They used a new concept of integrating the inverse of path width and defined a global law. To test their method for complex paths, they studied a new configuration called the spiral tunnel and designed with different path width and number of curves, and defined a local law. These laws have some limitations. For example, there are upper bound limits to the path width and the starting position influences a lot for the difficulty of a steering task. However, the paper still gave a lot of insights to quantitative design for HCI.
Ameya Daphalapurkar 10:22:39 9/20/2015
The paper '"Beating" Fitts' law: Virtual enhancements for pointing facilitation‘ is about the advancements and new research in the pointing respective of the virtual world and the real world. It states the different perspectives and principles with specific formulations that prove that by decreasing the target distance or target width or both can help beating the Fitts law in virtual world. An input device is the intermediary for virtual word pointing which is in contrast with real world physical objects. Some researchers have shown that intermediary free pointing in virtual world is possible with mouse or stylus like input devices. Further research also shows that beating the Fitts law is possible with certain amount of optimization in distance and the widths of the targets devices. The paper thus discusses on the background and various pointing facilitation techniques. Iterative correction model, impulse variability model and most optimal being the Optimized Initial Impulse variability model are some of the explanations for Fitts law. Achieving all this is all related to one thing and that is finding out the optimal balance in D and T. Control space, Visual space and the Control Display transfer function are the major factors affecting virtual pointing performance. Implementing different menus help to reduce the distance and make them equidistant as well from the cursor. Drag and pop helps with potentially bringing the target closer to the cursor. Skipping across empty spaces jumping from one selectable target to another is another method. Optimizing the area cursors but also dealing with the problems in large cursors and expanding targets are the other methods. **************************** The paper titled ‘ Beyond Fitts' Law: Models for Trajectory-Based HCI Tasks ‘ explores the existence of regularities in trajectory-based tasks with steering through tunnels as experimental paradigm. The Fitts law help us derive the fundamental logic that the logarithmic equation of distance and width help us derive the index of difficulty and which has the relation that greater the ID the more difficult is a task. Paper shows why Fitts law is not the optimal solution for all the devices and how trajectories like drawing is difficult with the mouse and simpler with a pen or stylus but still the values are close enough for both in Fitts law. Steering through tunnel is another experiment where the one has to draw a line from one side of figure to another passing through tunnel. Studies show that deviations are high if the movement is fast thus concluding that larger amplitude turns into less precise results. Experiment one consisted of just passing from Goal1 to Goal2 and the movement time was analyzed and it turned out to be agreeing with Fitts law. Next experiment is to increase the number of goals and thus it resulted to conclude that it is similar to the steering through tunnel task. Step included are passing from Goal1and2 and then increasing the amplitudes to see the current relation to be twice the logarithmic equation and process goes on for N steps. Narrowing the tunnel leads to higher number of error percentage. Spiral tunnel was conducted to check out for changing paths and it was confirmed that complex tasks also validate the prediction of difficulty. The overall conclusion is that there exists a relationship between time and tunnel width in steering and that time is linearly related to ID and speed is related to the number of constraints.
Manali Shimpi 11:16:53 9/20/2015
Beating Fitts' law: Virtual enhancements for pointing facilitation: It is shown that virtual pointing can be modelled using Fitt’s law. However virtual pointing is not constrained by the laws of physical world and can be improved artificially. In this paper a survey of recent researches to improve pointing performance is done. The researches are roughly categorized into 3 cases which are :1. Attempt to decrease D .2. attempt to increase W. 3. Decrease D and increase W.The author then discusses about the motor control models that explains the Fitt’s law in which optimized initial impulse model was more complete. Based on the models, author concludes that, in researches to decrease D should focus on initial large and fast movements and the one with decreasing W should focus on final corrective movement phase. In improving pointing by decreasing D, author discusses various methods. First is to design widgets that minimize D Even though some elements can be effectively transformed to reduce D , common elements like icons, menus are not easily transformed without complete redesign of the interface. Second is bringing potential target to the cursor that is simple drag and drop. However this method works fine in virtual sparse desktop but problems of occlusion and false activation appear in case of dense desktops. Third is removing empty space between target and cursor. The author discusses about the object pointing in this case. In case of facilitating pointing by primarily increasing w , the author explains two methods. First is area cursor where very small targets can be easily selected using area cursor than pointing cursor. Second is expanding targets in which widgets are dynamically expanded to a usable size only when required. In case of decreasing D and increasing w , there are two methods, one in which control- display gain is dynamically changed.----------------------------------------------------------------------------------------------------------------------------------------------------- Beyond Fitts' Law: Models for Trajectory-Based HCI Tasks: In this paper the existence of robust regularities in trajectory based interactions are explored where Fitt’s law does not apply The author further explains that if the movement is too fast then the small deviation from the trajectory results in constraints being exceeded before any feedback analysis or movement correction. Several experiments are performed to derive relationship between completion time and movement constraints. The first experiment was goal passing in which subjects were asked to pass goal 1 and goal 2 with two consecutive sets of 9 A-W conditions. The result showed that this task follows same laws as Fitt’s tapping task. The second experiment was by recursive analysis to formulate hypothetical relationship for steering task. The third experiment was made test if the method can be applied to linear trajectories with variable widths. In the fourth experiment, subjects were asked to draw the line from center to the end of the spiral. The result confirmed that the difficulty of steering task is also valid for more complex tasks.
Matthew Barren 16:04:57 9/20/2015
Summary of “Beating” Fitts’ law: virtual enhancements for pointing facilitation: Balakrishnan examines how researchers are optimizing pointing through decreasing distance, increasing width, or a combination of both. In doing so, Balakrishnan discusses how these pointing changes address three key factors motor/control space, visual space, and the control-display. In “Beating” Fitts’ law, Balakrishnan explains how there is an opportunity to provide improved pointing access in a virtual space compared to an analog environment. A virtual interface can be altered dynamically to better suit the user’s needs at given points during an interaction. For example, he discusses how Apple’s OS allows for the icons to change in size as the user scrolls over those objects. In particular, there is an opportunity for interfaces to change to decrease movement time and increase accuracy, but I believe a mac style expanding icons is not particularly useful. As Balakrishnan shows in his paper, the dimensions of the dock bar and icons change in size. This brings an issue of seeing an icon in one area, and then the icon quickly changing its current location when a mouse is scrolled over it. This occurs at a rate where it is impossible for the user to truly see the transition of the dock bar change, which requires the user to again review the position of the icon. Also, there is an overlap of icons, which can make it hard to select adjacent applications. After reading through the paper, there are some redesign ideas that will better facilitate pointing. One such idea is to have dynamically expanding icons based on access. The idea being, the icons you access the most will appear larger on your screen, and the icons that you access the least will be smaller. This would provide a discerning characteristic to icons based on the quantity of use and increase the width of more desirable objects. Additionally, rather than expanding an icon, a cursor could cause a change in color to an icon to show that it is currently over the target. This would provide feedback to the user to let them know the target they are currently over, and at the same time, eliminates the problems of overlap with expanding icons. In Balakrishnan’s paper, there is also a discussion on hotkeys and how this is a way to reduce the distance between objects and increase accuracy. Hot keys are implemented in many forms that most users only take advantage of an extremely small subset. One useful hotkey is to click the first letter of the action you would like to perform in a drop down menu. For example, a user can go to “File”, click the letter “c” on the keyboard to highlight copy, and tap enter when the key reaches copy. Hot keys are a great feature because of the ability to look at items, and not have to measure the distance of the cursor to that item. Instead, the user can remain focused on the item, and if typing is second nature, the user can quickly and accurately access the desired target. Summary of Beyond Fitts’ Law: Models for Trajectory-Based HCI Tasks: Accot and Zhai explore the dynamic motions of today’s target acquisitions on interfaces. They note that Fitts’ Law is limited in its ability to describe tasks which require a series of changing trajectories, dynamic widths, and amplitudes. In Beyond Fitts’ Law: Models for Trajectory-Based HCI tasks, Accot and Zhai, hypothesized a new procedure for determining difficulty and time to complete a task with trajectories. As they note, using an interface can have a whole host of different variables opposed to the typical open space pointing and clicking. They looked at steering a stylus through a set of varying sized paths. In some instance, these paths would dynamically change by shrinking or take on a conical form like the golden spiral. When looking at these tasks, they examined how the amplitude (length of the path) and width affected the time and difficulty with the task. The authors note that their method of examining steering is applicable to comparing menus where the user needs to make trajectory changes in order to reach a sub menu. In just a design point of view, accessing submenus based on cursor pointing can be quite frustrating. The speed that can be achieved with a mouse makes it challenging to get the desired accuracy to reach a target. In their procedure of experimentation, it would have been interesting to see an attempt at optimizing speed of a cursor at varying points of amplitude and width. For example as the distance from the beginning increases and the width stays the same, it may be helpful for the cursor to decrease the available speed. Also, as the width of a path narrows, the speed of cursor could become less sensitive to avoid a loss in accuracy. This same idea can be extended to drop down menus. As a user enters a submenu, the sensitivity of the cursor can decrease to avoid falling off of the submenu and losing the target. Additionally, the submenu could widen in size, making it a larger target than the previously accessed menu. Again, this would provide another way to increase the accuracy in this type of motion.
Kent W. Nixon 17:47:53 9/20/2015
Beating Fitts' law: Virtual enhancements for pointing facilitation This reading discussed ways in which the upper bound of Fitts' Law can be overcome in virtual interfaces. The possibility to do so exists as unlike the physical world, the virtual world can be manipulated in realtime along with user input in order to make objects more easily selected. Namely, the distance, D, and width, W, features can be manipulated by either scaling or moving objects onscreen. The paper also differentiates between visual distance and width, and control distance and width. The former are manipulated via the previously stated methods, while the latter can be artificially scaled via forms of pointer acceleration, sticky objects, and etc. The paper is interesting to mean in that it differentiates between visual distance and control distance. I would be interested to see if there was some sort of mathematical relationship between the two – for example, an extended Fitts' Law. Beyond Fitts' Law: Models for Trajectory-Based HCI Tasks This paper discussed a quantitative model for estimating task completion time when moving a cursor along a path of defined length and width. Starting from Fitts' Law and data collected for a “goal” game, the authors derive task time models for various tunnel navigation tasks, and show that their calculated models match closely with experimental evidence. I found this paper to be extremely interesting as the authors were able to achieve a surprising accurate expectation of experimental results simply through mathematical derivation of various models. I am also studying some similar mathematical techniques in a quantum mechanics course, so that was cool, too.
Mingda Zhang 17:58:58 9/20/2015
"Beating" Fitt's Law: Virtual Enhancements for Pointing Facilitation The first recommended reading is a review paper published in 2004, which summarizes the efforts of modern input device trying to improve users' pointing and selection experiences. Although ten years have passed since then, some ideas from this systematical study are still valuable as guidance of current design. This paper first reviews the famous Fitt's law and its original representation: an empirical estimation of physical movement time from the distance and size of item. Although Cards et. al. have validated that Fitt's law is also followed in digital world, people are still trying to improve the performance of various input device. In fact, since input device plays the role of intermediary in human computer interaction, it is inherently possible to bypass the restriction of real world physics law. According to the most widely accepted theory, the underlying motor model behind Fitt's law is simple: people's movement for achieving targets can be interpreted in two phases: first an initial muscle impulse rapidly gets subjects close to the target, and then a closed-loop feedback control helps subjects to accurately hit the target. Corresponding to Fitt's law, the first phase is mostly responsible for the factor of distance and the second phase is for target size. Similarly, people have made progress in reducing distance or increasing the target size, or both. To reduce distance, people have tried to use new layouts to replace linear lists, or temporarily bring relevant targets closer, or simply make cursor jump from one potential target to the next (which is totally impossible in real life). Although significant progress have been achieved in controlled experiment, the performances of these approaches are mostly dependent on accurate predictions of candidates. Without this assumption, some techniques can be troublesome rather than beneficial. Approaches to increase the size include using area cursors(larger cursor) and dynamically expanding targets(larger target). These techniques are theoretically effective and are proved to be useful in selecting isolate targets, but neither of them could satisfactorily handle the clustered targets. Obviously, more sophisticated strategies need to be introduced to help handle the accuracy problem, which would inevitably cause problems in seamlessness. Another promising solution of the problem focuses on both directions, trying to altering the ratio of movement distance. In other words, when human moves the mouse (or other input device) rapidly, the underlying assumption is that user is trying to cover large distance, thus the cursor would move faster. This is pretty similar with the design of some modern vehicles: the steering of wheel under different speed has different effect on vehicle direction. However, this approach is also suffering for selecting closely aligned tasks as most of the case in daily use. As a validated quantitative rule, Fitt's law plays key roles in evaluating performances of different input devices. More importantly, it provides insight to improve the performance. Although the limitation of human themselves restrict the ability to reach real life objects, it is not necessarily true in digital world. Beyond Fitt's Law: Models for Trajectory-Based HCI Tasks This paper seems more interesting because it learns from Fitt's Law and use thought experiment to expand Fitt's Law. Of course, the real psychological experiments finally validated their assumption. In Fitt's law, the key movement is tapping or hitting, which only cares about the starting point and the goal. The authors assumes that more complicated task, like constraint movement or trajectory-based tasks, should also be governed by a similar rule. The underlying hypothesis is based on a thought experiment: if adding more and more constraints to the tapping experiments, the difficulties would increase and to the extreme, a tapping experiment with infinite constraints is essentially equal to a trajectory-based task. This is a significant expansion of Fitt's law because it tries to solve a more sophisticated but more frequent tasks in graphical user interface design. Specifically, in modern computer interfaces, few tasks include only moving and pointing. The psychological experiment proved the hypothesis that a steering law, according to the authors, are more suitable to describe the task involving producing trajectories. The authors also designed experiments with variant width and moving directions, thus further formulate a more general rule. This quantitative discovery expanded the arsenal of interface design theory and provided more possibilities in better design.
Chi Zhang 22:58:06 9/20/2015
Critiques on “Beating Fitts’ law: virtual enhancements for pointing facilitation” by Chi Zhang: This paper is a survey on recent research about target pointing technologies. There are many methods introduced in this paper about reducing the target pointing time of graphic user interfaces by artificially reducing the target distance, increasing target width or both. This is the key to the speedup of movement, which is quickly noted by the author. As we can know from this paper, Fitts’ Law is actually model of human movement, predicting the time of rapidly moving to a target area. However, there are problems with targets being too close, as the calculations would be inaccurate. This is a very good survey paper, it presents all recent researches about target pointing techniques and also gives out concerns and future problems. ----------------------------------------------------Critiques on “Beyond Fitts’ law: models for trajectory-based HCI tasks” by Chi Zhang. As we know from previous paper, Fitts’ law determines the physical law of target pointing tasks. However, Fitts’ law cannot be applied into trajectory-based tasks, like navigating through nested-menus, drawing curves, moving in 3D worlds. The authors novelly created a notion called “steering law”. They consolidate their law by using four examples. First, the authors set up a simple goal to pass task to establish the new model. Second, they set up a steering task based on the model. And after these, more tasks are designed with complexity in order to derive the final model. It’s a very interesting paper as it presents new idea upon problems that needed to be fixed in real world.
Priyanka Walke 0:08:46 9/21/2015
Reading Critique on Beyond Fitts’ Law: Models for Trajectory-Based HCI Tasks The author has explores the robust regularities in movement based tasks especially trajectory based tasks. The exploration also includes the derivation of a global & local law in order to identify the relation between time taken and the velocity of movement with the width of the path. The paper mainly focuses on the experiments conducted to model the trajectory-based tasks for which the Fitts’ Law proves to be insufficient in case of pointing to steering. The experiments that were conducted, mainly consisted of Goal Passing, Goal Passing with larger number of constraints, Narrowing Tunnel & Spiral Tunnel. Every experiment was conducted to measure the change in movement time with respect to the amplitude & the width of a trajectory-based task. The experiments stated that as the width decreases, the error rate as well as the time increases, leading a conclusion that they were linearly dependent on the index of difficulty. In case of a simple linear relationship between time and tunnel width, all the experiments gave a correlation greater than 0.96 in case of steering tasks. The Fitts’ Law limits itself to the drawing tasks and hence it is important to move beyond and understand other relations as currently we are dealing with much complicated stuff like, 2D, 3D pictures. Also, the findings stated in the paper are only precise up to an upper bound limit on the width for the laws model the relation correctly, which is definitely limited. The menu selection shows the design consequences of the findings mentioned. The menu also includes how to measure the time in that case which depends on the vertical/ horizontal motions, i.e. the number of items and the width of the menu. However, there are a few things that need to be considered which include the Pointer Width, Speed, the C-D gain & its effects on trajectory-based tasks while selecting as they take more focus on pointing. Also, it’s the Speed that leads to most of the errors than the width. The length just mentions the time consumed. Hence, it can summarized by saying that much more research is possible in this field. The authors emphasize on the importance of moving from pointing to steering based tasks, and derive the laws to show that the movement time & velocity depend on the path width in such conditions. Reading Critique on “Beating” Fitts’ Law: Virtual Enhancements for Pointing Facilitation This paper involves a detailed discussion of all the techniques & different approaches to interface design that focus on ‘Beating’ the Fitts’ Law. The paper covers a larger area of discussions made by numerous researchers since long time. It first explains the Fitts’ Law as the effect of distance and width of the target object on the movement time. Also, it refers to the distance & width as the motor and visual spaces. The human selection motion remains as it is, as a singular initial impulse suspended on the limb towards the target object and a closed loop feedback. Lot of different techniques have been used which make use of hotkeys, scrolling wheels, use of proxy images & Object pointing for decreasing distance. For decreasing width, area cursors & expanding targets were used, of which area cursors were highly successful. The problem that persists throughout is that of densely populated displays. When faced with the densely populated displays in case of C-D gain it’s the presence of the involving objects, which itself is due to the densely populated displays. A comparison between these techniques & Fitts’ Law is not possible which complicates the situation of choosing a better technique. The paper covers huge of amount of research experience of many researchers, which itself is a rich source of information. The paper suggests about a couple of important concepts which can be applied to some situations. The auto-correction technique which states that you do not have to be over the object, being in the vicinity should redirect you to your target. The problem that we have to deal here with dense population and much similar to a C-D gain problem. We can think of a screen with an anchoring system like mac on all 4 sides of the screen. Hence, there will be no icons on the screen, thus maintaining the workspace clean. We can select a combination of expandable files, proxy images, C-D gain to create this. The techniques mentioned above can be applied to this.
Shijia Liu 0:18:29 9/21/2015
Section 1: "Beating"Fitts' law:virtual enhancements for pointing facilitation. First of all, the Fitts law means that predicts that the time required to rapidly move to a target area is a function of the ratio between the distance to the target and the width of the target. According to the article, we could find that there is a formula is very useful for us to understand this is : MT=a+blog(D/W+1), which indicates two possible approaches for further optimization: reduce D or increase W. Actually the challenge is to indirectly affect further changes in D or W in ways that do not substantially alter the overall visual appearance of the graphical interface, but nonetheless result in shorter pointing times. And in this article, the author also talked about the 3 various pointing facilitation techniques that have been developed by now::(1) primarily attempt to decrease D. (2) primarily attempt to increase W, and (3) both decrease D and increase W. The most successful and complete explanation to date,called the optimized initial impulse model, it is kind of the hybrid of iterative corrections model and the impulse variability model. In additional, there are three major factor that come into play and can affect performance in virtual pointing: motor or control space(Dm,Wm), visual or display space(Dv,Wv) and the control-display(C-D) transfer function that links the two spaces. According the 3 various pointing facilitation techniques: For the first one: Facilitating pointing by primarily reducing D, it also has 3 subsections in it: (1) Designing widgets that minimize D: in the article, the author give us a very simple instance to analyze how to minimize D without change altering Dv or Wv, but not every case we could use this simple graphic path to accomplish it., other common interface elements such as icons, buttons, and scroll-bars are not easily transformed to reduce their distance to the cursor, without drastic redesign of the overall interface layout. (2) Temporarily bringing potential targets to the cusor, for example: Drag-and-pop. The overall interface is unchanged, but the user can now manipulate the proxy icons at a much closer distance. However, it still has defections, first, the proxies do not overly clutter the region near the cursor or obscure valuable information around it,furthermore, it can be tricky to implicitly determine when the user intends to select the remote elements versus items that are already in the nearby vicinity .(3) Removing empty space between the cursor and targets. There is an interaction technique, called object pointing, where the cursor essentially skips across the empty apce ,jumping from one selectable target to another. So, Dm is reduced and Dv is unchanged. However, there still have a few practical defects that may lower its overall applicability and value, and thus hinder widespread adoption. For the second one: Facilitating pointing by primarily increasing W. (1) Area cursors:this technique has significant problems with this issue: large area cursors can obscure underlying data, and it can be difficult if not impossible to use area cursor to select one target from several targets closely grouped together. (2) Expanding targets: the size of the interface widget or viiewing region dynamically changes to provide the user with a larger target area to interact with at their focus of attention. But it also has counterexample to the benefits of expanding targets is indicated by Gutwin's study of target selection in a fisheye view. For the third one: Facilitating pointing by both decreasing D and increasing W:(1) Dynamically changing the control-display gain: The ratio of the amount of movement of an input device and the controlled objects is refereed to as the C-D gain. Target aware C-D gain adaptation undoubtedly improves pointing performance for single isolated targets, problems arise when multiple targets are present in that intervening targets act as traps that can hinder movement along the way to the desired target. As all the technique above, obviously, each of them could enhance the performance, however, at some extent, all of them have some defections as well. While the techniques tend to produce significantly improved pointing performance when selecting isolated targets, difficulties arise when they are used for selecting of multiple targets that are spatially close together. Furthermore, it would be worthwhile to conduct further studies comparing these various techniques to one another directly. And the best measure of all is elicits a comment from users. Section 2: Beyond Fitts' Law: Models for Trajectory-Based HCI Tasks Concerning the last article, we could got a consideration is: The Fitts' Law somehow users' performance in the tasks can't be successfully. Therefore, in this article , a new law coming up is called the steering law, and the author found that the Steering Law indeed exist by 4 experiments( Goal Passing, Increasing Constraints,Narrowing Tunnel,Spiral Tunnel). And with the Fitts' Law performance scores(pointing/tapping times) can be translated into a performance index that is independent of those experimental details. The value of Fitts' Law is significant. However, nowadays, computer input devices are used not only for pointing to targets but also for producing trajectories, such as in drawing, writing, and steering in 3D apace.Fitts' Law is not suitable for them now. In this research, the paradigm is focus on is steering between bound-aries. They took several steps to see the quantitative relationships between completion time and movement constraints in trajectory-based tasks. In the first experiment ( Goal Passing) it shows that the movement time against Fistts' ID shows a linear relationship with a high correlation between them. In the second experiment( Increasing Constrains): it shows that there is a strong correlation between the hypothesized model and the data collected like the chart shows in the article, the time and the index of difficulty is proportional. In the third experiment ( Narrowing Tunnel): It kinds of another proof of the successful trails and index of difficulty for the narrowing tunnel steering task forms a linear relationship( proportional). In the last experiment( Spiral Tunnel): It also the same issue like the 2nd and 3rd experiment did. So the author carried the spirit of Fitts' Law a step forward and explored the possible existence of other robust regularities in movement tasks. Furthermore, the intergral form states that the steering time is linearly related to the index of difficulty, which is defined as the integral of the inverse of the width along the path; the local form shows that the speed of movement is also linearly related to the normal constraint.
Zihao Zhao 0:24:04 9/21/2015
The article ‘‘Beating’’ Fitts’ law: virtual enhancements for pointing facilitation is a survey which discusses the difficulties and challenges in user graphical interface. How to enhance the performance of pointing in user interface is the main problem that has been discussed. According to Fitts’ law, there are three ways to enhance the performance of pointing, reduce the distance between the cursor and the icon, make the width of the icon bigger, or apply them at the same time. Dynamic width of the icons is very useful in enhancing the pointing performance. Apple OS has applied it in it’s system and the dock is a line of icons that provide the shortcut for users. When the number of icons become bigger, the size of the icons will be reduced automatically. Consequently, there are two main advantages by applying it. 1st is that it largely minimize the blank space so that users can point to the target more efficiently, when the number of the icons reduces, the size of the icons will become bigger and the moving time for user to get to the target will be reduced. 2nd is that it prevent the situation that icons group together when there are more icons on the dock. Although the dynamic icon width is very useful theoretically, but the excessive use of it will affect the arithmetic of the design in graphical interface.———————————————————————————————————— The article Beyond Fitts’ Law: Models for Trajectory-Based HCI Tasks mainly analyzes the drawback of Fitts’ Law and carried Fitts’ Law a step forward. Fitts’ Law is a law in human- computer interaction that measures the performance of the techniques. However, based on the time and the history it was carried out, although it is very useful in graphical interface analysis, it can not be applied to some new input and output relations of human computer interaction. For example, 3D inputs and the electronic helmet which give the users the feeling of reality. The article explored the possible existence of some robust regularities in movement tasks, so that we can generate some mathematical relations in new inputs like Fitts’ Law to solve this problem. In my point of view, this may be useful in some circumstance, but it is not vital in the measurement of some interface design. Because the way of human computer interaction is too diverse and it is changing drastically. Like the mobile devices, the spirits of Fitts’ Law can seldom be used in the assessment of mobile computing because there is not any mouse for the controller and it’s just gestures. So there is meaningless to discuss the extension of Fitts’ Law.
Lei Zhao 1:10:58 9/21/2015
The first paper talks about the state-of-art techniques on improving pointing performance in GUI interfaces. First, the paper describes the definition of Fitts' law by introducing each component in the formula. In this part, the author focues on explaining why Fitts'law in virtual world differs from the physical world. Then, the author groups different existing techniques into 3 categories: 1) techniques aims at dicreasing D, 2) techniques aims at increasing W, 3) techniques aims at both D and W. The main contribution of this paper is to evaluate all these three kinds of optimizations. For decreasing D, although it can reduce the distance of target, it also makes the selection of originally near target more difficult. For increasing W, it may cause a problem of make the interface design ugly by making the icons too big. At last, the authors get a conclusion that all these techniques have the same problem: it is not very efficient when selecting multiple targets. The second paper is called "Beyond Fitts' Law". This paper talks about a steering law for trajectory-based task. In this paper, three experiments are conducted to investigate how the completion time relates to the movement constraints in steering trajectory-based task. According to the experiment result, the authors generalizes a global law that can predict the total time to perform a steering task. The main idea of the law is: the velocity is in proportion to the width of tunnel, and is inversely proportional to time constant. In my oppinion, the conclusion of this paper, especially the equation generated by the this paper, gives a very good opportunity to extend for further research and used for quantity assesment.
Samanvoy Panati 1:18:19 9/21/2015
Critique 1: Beating Fitts’ law: Virtual enhancements for pointing facilitation This survey talks about various approaches and experiments conducted for improving performance in pointing at targets in graphical user interfaces. It talks about beating Fitts’ law with the help of computer by changing the variables in the virtual world. Virtual pointing can be modeled using Fitts’ law as MT = a+blog2(D/W+1) where MT is the movement time, a and b are empirically determined constants, the logarithmic term is called Index of Difficulty(ID) and reciprocal of b is called Index of Performance(IP) The techniques developed to facilitate pointing can be divided into 3 groups. 1) Decreasing D 2) Increasing W 3) Both decreasing D and increasing W The survey gives the introduction about iterative corrections model, impulse variability model and their hybrid optimized initial impulse model. The standard deviation of the endpoint of any movement increases with the distance (D) covered by that movement, and decreases with its duration (T). S = k(D/T) where k is a constant. So long distance and short duration movements result in high standard deviation and thus increases the error rates. Now the survey gives experiments, approaches, advantages, disadvantages and future works in every category. 1) Reducing D: To reduce distance and improve movement time, linear pop-up menus can be used so that more data can be made closer and available in a single interface whenever required. Pie menus give an example for equidistant options in a menu. Another approach is to bring potential targets temporarily closer to the cursor. The third approach is to reduce empty space between cursor and targets. These spaces normally constitute more than 90% of the pixel space. The spaces are given for better view and usability. The spaces can be skipped. One usage of that is using Tab to skip the empty space. Thus Dm (motor space) is reduced by keeping Dv (virtual space) constant. However, these approached depends on target density. If many targets are closer to each other, then these are of not much use. They may lead to false triggering and so result in higher error rates. 2) Increasing W: The first approach is Area Cursors where instead of increasing the width of the targets, the width of the cursor is increased. This may lead to problems like obscuring the targets which can be solved by increasing the cursor area only when it is required and using the normal one when not required. The second approach is to expand the targets. When the cursor is moved over the targets, the targets can be expanded thereby increasing the width of the targets. Further, these approached suffer with their performance when the targets are close to one another, that is, when the target density is high. 3) Decreasing D and increasing W: The C-D gain is the ratio of the amount of an input device and the controlled objects. The first approach is to use ‘sticky targets’ where the cursor slides over with high speed when there are no targets but moves very slowly when it is on the targets. The problem appears when there are many targets between the cursor and the required target. Also, the buttons used more frequently and that are certain to be clicked are made much larger than the others for easy usability. These approaches are very important to know because there is a lot of research going on in these areas which can significantly improve the usability of the interfaces. In the survey, many approaches are measured based on error rates and other quantitative measures but they can also be measured based on the acceptability of a technique by the end users which will be more efficient parameter. ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ Critique 2: Beyond Fitts’ Law: Models for Trajectory-Based HCI Tasks This paper explores the possible existence of robust regularities in trajectory-based tasks. This explains about the four experiments conducted to analyze the relation between different parameters and finding a simple “steering law” for trajectory-based tasks. It firsts gives the importance of Fitts’ law in human computer interaction and how it was used for various comparisons between the interfaces. The technology is so pumped up that now input devices not only point to targets but also produce trajectories such as drawing, writing and steering in 3D space. Fitts’ law is not enough for these trajectory-based tasks. The following experiments are conducted in research. Experiment 1: Goal Passing The subjects were given a stylus and stylus and a tablet. The task is to pass Goal 1 and Goal 2 without touching the boundaries and lifting the stylus. The results show that there is a linear relationship between the movement time and Fitts’ ID. MT = -1347+391log2(A/W+1) where A is amplitude (for distance D) Experiment 2: Increasing Constraints In this experiment the goals are increased gradually. The amplitude is divided in every step. For n steps, the amplitude became A/N. N is extended to infinity and this is called “tunnel travelling”. Again the tests are conducted and the resulting expression is MT = a+b(A/W) After subjecting the people to experiments the resulting movement time is MT = -188+78*ID The error rate increased significantly when the task becomes very difficult. The average error rate is 6.4% Experiment 3: Narrowing Tunnel In this experiment, the width of the tunnel is not constant. It decreases gradually when getting closer to the goal-2. The result is MT = -532+93*ID Since the width is not same, the average error rate is around 18%. When the same approach is done with the curvy tunnel to establish a general formula, IDc = integralc(ds/W(s)) Experiment-4: Spiral Tunnel Here the tunnel is in spiral shape. The path should be drawn from the center to the end of the spiral. The width is minimum at the center and maximum at the end. The subjects are subjected to this experiment and it confirmed that the prediction of the difficulty of the steering tasks is also valid for this complex task. The time to steer through the spiral path is linearly related to the index of difficulty. MT = 115+169*ID The average error rate is 13.7%. A local law is derived, v(s) = W(s)/r where v(s) is the velocity of the limb at the point of curvilinear abscissa, W(s) is the width of the path at the same point and r is the empirically determined time constant. The performance in these experiments also depend on some important behaviors of the subjects like whether steering is performed from left to right or right to left, clockwise or anti clockwise and whether the subject is left handed or right handed. Using these relations and formulae, many design tasks can be compared and then decision can be made on which one is more productive and useful. This information can be used as a means to compare designs.
Zinan Zhang 3:43:22 9/21/2015
1. For ‘‘Beating’’ Fitts’ law: virtual enhancements for pointing facilitation. ------ This paper mainly talks about finding a new law to apply to human-computer interaction research and design. Based on the very few robust law, Fitts’ law, the authors use four experiments to discover a new law’s existence. Finally they find some important relationships, which can enrich the future research in the field of HCI. Although the authors did not get a newer and better law than the Fitts’ law, their method of researching is still a reasonable approach I think. Every step they take is because they find the deficiency of the prior method. When they figure out deficiencies, they try to make a assumption about it and then design a little experiment to approve it. With the help of their collection in the experiment, they try to figure out a relationship between the two variables in the experiment. The relationships are quite useful for the authors to find the new laws. In addition, if I am going to do this research, I will add one more experiment. The types of the tunnel, besides straight, narrowing and spiral, should add a right-angle tunnel. The trajectory looks like a angle of a rectangular. Because in most time people uses computer are inevitably to select objects on the top menu bar (always pop-up linear menu). And that require the users to select with a track of broken line, which is not convenience for users. This case should also be considered in the new design laws. ------------------------------------------------------------------------------------------------------- 2. For Beyond Fitts’ Law: Models for Trajectory-Based HCI Tasks.------- This paper mainly also talks about finding a new law to replace the Fitts’ law. The paper introduces several attempting to create a better law for HCI, and groups the attempting in three categories. The first group is by reducing the distance between the cursor and the target. Above all the method mentioned in this group, designing widgets that minimize the distance between cursor and target is the most reliable method I think. It is the most possible way to come true. Perhaps some days in the future, people will design the menus like a pie and it will pop-out when people want them to show up. And that will be significantly easy for users to next work. The second group is to increase the size of the target or the cursor. One of the examples of increasing the target’s size is the Apple MacOSX dock. The icons in the desktop toolbar expand when the cursor is over them. As a user of Mac, I think it is a great design. With the help of this design, it is so easy for me to click the icon I want every time. When I use the windows operating system, sometimes I will click a wrong icon because they are just so close and tiny. As for the last group: both decreasing the distance and increasing the size, I think that is a good idea but not easy to come true. Because there are two objects the designers need to care about. And that is too much for a designer to think about during the designing. So in my opinion, designers should focus only on either decreasing distance or increasing size. That is much more possible to work out a better design.
Darshan Balakrishna Shetty 7:09:19 9/21/2015
1. “Beating Fitts' law: Virtual enhancements for pointing facilitations”: This paper mainly discusses about all the recent developments in the technology of pointing at targets, how can we supercede the limit of Fitts' Law which gives the relationship between the time taken to point to the target with respect to the distance to the target and the width of the target. Even though the paper discusses many methods to improve the performance keeping in mind to beat the Fitts' Law but methods discussed in the paper all base there theory on Fitts' Law. This shows how significant and important Fitts' Law has been. Based on the Fitts' Law the performance of the pointing device can only be improved by 3 methods i.e., by decreasing the distance between the source point and the target, increasing width of the target or we can also say as discussed in the paper by increasing the width of the pointer as well and the last method is the combination of both decreasing the distance and increasing the width of target/pointer. For each of these methods the we can see that when there are too many icons there is no much performance improvement rather in most of the cases the methods discussed deplete the perdormance. Of all the methods I liked the method in which area of the pointer/cursor was increased and the cursor also had a cross which has a single pixel point in between the area of pointer. So that it can handle both the cases like for few icons in which if any part of the cursur touches the icon it is selected, and also in case there are many icons cluttered and the area of cursor covers a number of target then whichever icon is touching the cross point is selected. Nowadays if we say the performance of the pointing facilitations are improved on the capacitive touch screens where the user uses his fingers to select the menu or icons which is more direct that the visual and motor co-ordination which was needed earlier between the screen and the pointing motor device. All in all it was an important paper to read the significance of the Fitts' Law. 2. “Beyond Fitts' Law: Models for Trajectory-Based HCI Tasks” : This paper discusses about the short comings of Fitts' Law. Fitts' Law mainly deals with a simple linear target at a distance D, with a width of W but what if the user has constraints of going through a predefined trajectory say as in a game or if the user using paint application where he has to draw kind of trajectories etc, then what would be the performance/ time equation. The author tries to come up with a equation in these scenarios using the Fitts' Law. The author models 4 experiments to come up with the steering law/equation. In the first experiment author tries to set up a tunnel, where the user has to draw a line between 2 barriers kind of a tunnel. The author starts with 2 simple goals which the user to pass on similar lines he increases the number of goal lines and using the Fitts' Law comes up with an equation for the tunnel. Similarly the author sets up the model for tapering tunnel and spiral tunnel to make it a general expression. Even though the paper says going beyond Fitts' Law the paper is more of generalising the Fitts' Law for a trajectory based human interaction. Again the paper is more of a proof that how significant and important Fitts' Law has been in HCI.
Ankita Mohapatra 8:03:06 9/21/2015
Reading Critique on Beating Fitts’ Law: virtual enhancements for pointing facilitation: The primary argument for attempting to “beat” Fitts’ law (really, take advantage of our understanding of it) is that pointing and clicking is an extremely common task in the age of graphical user interfaces, thus any small enhancement in the speed of that action could theoretically result in huge gains in user productivity. The authors argue against the common conclusion that since mouse pointing is as good as physical pointing, it must be optimal in its current form. They suggest that since virtual pointing is not under the limitations of the physical world, there may be ways to enhance mouse performance and reduce movement time. The authors introduce a model of the motor control models behind pointing called the “optimized initial impulse” model (Rosenbaum), describing the process of pointing as one initial movement toward a target, followed by smaller corrective movements if the target is missed. The key component of the model is an equation for standard deviation of the endpoint, S = k(D/T), where D is distance and T is duration of the movement or time. A distinction was made between motor distance and virtual distance. An example of where this distinction is important is in the addition shortcuts on input devices (keyboard macros for buttons, mouse scroll wheels for scroll bars) that eliminate motor distance while virtual distance still exists between the cursor and the target (icons or scroll bars). The methods surveyed that only focused on reducing virtual component distance were disappointing. I felt like all of them were disruptive to the aesthetic of the interface and impractical/unusable. The exploration of width-focused methods were, on the other hand, really interesting, and features I have witnessed in practical user interfaces. After the survey of those W-focused methods, the authors concluded that the use of the dynamic properties of the cursor has a strong potential to enhance design interaction techniques. The overall conclusions made by the authors at the end of the paper were actually quite inconclusive in comparison. One thing I found surprising was how little attention was given to C-D gain adaptation in response to mouse acceleration, given that it’s a common feature in modern pointing input devices (e.g. mouse, trackpad), although this is a decade after the writing of the paper. It was considered a limited approach by the authors, who introduced it but immediately moved on to how it could be enhanced by knowledge of the objects on screen and users’ intended target. The authors focused very much on new techniques and only in the conclusion addressed the issue of user acceptance of the techniques. Many of the techniques involving gravitating the cursor toward nearby objects struck me as something that would feel disorienting and take away a sense of control. In general, I also felt that methods for reducing movement time focused only on reducing distance seemed extremely disruptive to the visual aesthetic of the interface, making them of limited practical use.===================================================================================================== Reading critique on Beyond Fitts’ Law: Models for Trajectory-Based HCI Tasks (1997) Jonny Accot, Shumin Zhai This paper set the lofty goal of trying to expand the body of quantitative methods that can be applied to measuring performance in interfaces on input devices. I was amazed at how much complexity was condensed into this paper in describing what they accomplished and contributed to future HCI research. The premise for the research was that Fitts’ law was one of the only quantitative methods for evaluating interface design, and it only modeled tasks involving moving over a single, straight-line distance to reach a target. What was accomplished was the discovery of a simple model for “steering through tunnels,” the authors’ analogy to trajectory-based navigation in both 2D and 3D, that was derived more or less from the principles of Fitts’ law. As much as I was amazed by the paper, I thought it was off to a weak start. They chose the two parameters amplitude and variability as inputs to a model of movement time when steering an input device between two boundaries. Although the concepts were very well explained, I thought that these were awful names for the parameters. I think the word “variability” describes more accurately what was called amplitude (I relate amplitude to the amplitude of a wave and imagine it as tangential to the direction of motion). If trying to describe the width of the tunnel (presumably without using the word width since it would conflict with Fitts’ law), why not a word like height or diameter to fit the analogy? If the desire was also to avoid the term distance, why not length or, even more accurately, displacement? Does it make sense to describe the diameter of the tunnel as “variability” even when it is not conceptually a rate of variation? Frustrating the reader further, the variable W was used for width (instead of V for variability === width) in their subsequent equations, and so there was inconsistency. The new models that resulted from their experiments were the magic of the paper. The authors discovered that multiple consecutive movement tasks could still be modeled correctly with Fitts’ law. They applied mathematical induction to generalize to the case of infinite discrete movement tasks, which is interpreted as the drawing of a path. They used further experiments to justify models for navigating through straight tunnels that changed in size, curved tunnels, and spiral tunnels. An interesting property of these models is that the index of difficulty is linearly, rather than logarithmically, related to A/W (“amplitude”/width) in all of the cases. That is, the discovered model for trajectory-based movement was MT = a + b(A/W).
Jesse Davis 8:27:39 9/21/2015
Beyond Fitts’ Law: Models for Trajectory-Based HCI Tasks This paper looks to extend Fitts’ Law by projecting (pun intended) the basis for Fitts’ Law onto a Trajectory-Based model. They do this by running several different experiments: the first one being a simple goal passing test, the next being a repeated goal passing test (i.e. the first experiment with several goals, one after another), the one after being a goal that’s narrower than the starting point, and the final one being a curved path towards a center goal. Each experiment brings the paper closer to deriving a “local law” based on their global law used to calculate the total time to perform a given steering task (each of the experiment is setup to be a steering task). The law holds well for all task variations with a little variation for the beginning of task 3 as can be seen in Figure 14 of the paper. The paper goes on to elaborate why this is an important study and how we can implement it in the world of HCI (menus, drawing, etc.) I found the paper very interesting and it sparked a few ideas for the basis of a master’s project. The figures and graphs were extremely helpful in giving context for the equations and without them it would’ve taken significantly longer for me to figure out what most of the paper was talking about. Beating Fitts' Law: Virtual Enhancements for Pointing Facilitation The previous paper (Beyond Fitts’ Law) looked to extend Fitts’ Law by looking at Trajectory-Based HCI Tasks, as explained, and this paper takes a different approach in utilizing Fitts’ Law. It seeks to virtually manipulate the variables in the equation. In particular: W and D, the width of a target and the distance between targets respectively. The goal is to optimize human-computer-interaction by either decreasing D or increasing W and they take a look at all permutations of this technique (reduce D, increase W, both). In the first section, I found the pie menu to be the most interesting, because I see it implemented already in several modern day games. I also thought that 3.1 could be utilized in 3.2, perhaps by rearranging the potential targets in a pie manner around the cursor (temporarily, while disabling the background, thereby temporarily creating a new temporary GUI in order to avoid interacting with the backdrop; if we use the desktop example from figure 3: we disable the icons on the desktop, we arrange the icons that can interact with the report in a pie formation around the report so that the user is able to drag it to whichever they desire; we could also perhaps use “likelihood to be used with this application type” and surround the selected application with different tiers of circles, with the inner most tier of the pie circle containing the most commonly used applications that interact with the application/file that the user has selected). The idea of area cursors was an interesting one, but doesn’t seem very intuitive, while the expanding target design is already implemented on macs (or at least it is similarly implemented). The attempt to use both is very promising, however it requires data analysis and it would work best as an adaptive model. Awesome paper, look forward to hearing about it in class.
Sudeepthi Manukonda 8:48:46 9/21/2015
The first paper, Beating Fitt’s Law talks about the importance of targets and movement towards the targets. Pointing and moving towards the targets is an important part to perform any action successfully. Therefore performance of any action also includes the time taken to move to the target. Thus to improve the optimality the time taken to reach the target has to be minimised. Fitt’s Law gives the descriptive model of the time taken to reach the target in the real world. Research work is going on in to reduce the time further in the computer world. There are two factors that affect this time variable and they are the distance to the target and the width of the target. the ratio between these gives the time. There are thus three types of optimising the time, reducing the distance, increasing the width or both.This paper talks about beating the Fitt’s Law. For a common human being, using a mouse might not be very different from the action that he might have performed in the real world. This paper deals with beating the Fitt’s Law and making the virtual pointing even better. Iterative corrections model and impulse model both put together give the most successful Fitt’s model application till date. Because the objective is to minimise the time taken, the former model assists in going to the target and if the target has not yet reached then the second model assists in clinching to the target. This paper talks about motor space, display space and the transfer that links both. It is called as (C-D) transfer. The situations are studied, factors are pointed out, and what has to be done to optimise has been laid out. Yet, there is no model that would perform a successful operation or give an accurate outcome in every situation. There are many ways in which this optimisation can be achieved. This is highly relevant to us because this is so much related to the usual human interactions with objects. There are targets for almost every action. The paper concludes that, the current research is not sufficient and it is yet to grow. There are several situations that are encountered in typical graphical user interfaces that require much more understanding and optimisation techniques. Such situations include pointing towards multiple objects, or moving to multiple objects. A partial solution to one such problem is to choose an optimal input device that can better assist in the given scenario. The second paper, Beyond Fitts Law, deals with the trajectory based interactions which include accessing nested drop down menus, movements in 3-D worlds, etc. Fitt’s Law has been the closest to the most optimum solution one had found to solving the distance problems in Human Computer Interfaces for its accuracy and robustness. But it is found that Fitt’s Law has address only one type of movement and definitely not trajectory movements. The paper first talks about Goal Passing. It talks about an experiment that was conducted in which two vertical segments were presented on the screen in green colour. When the stylus was pressed against the screen, there was a blue line that was indicating the trajectory movement. The colours kept changing with the movement of the stylus. But the results are the same as Fitt’s tapping task. Secondly, it talks about increasing constraints. This includes adding some constraints on both the ends to further test goal passing. But this is similar to the above too. But it was successful in describing the difficulty of the task. There are a couple more experiments that are performed namely, narrowing a tunnel and a spiral tunnel. The aim of this paper is to tell that the Fitt’s law cannot be successfully applied to the trajectory movements and to find such a law that can be applied. And by the end of the paper, it reveals that such a law exists called the steering law. All the experiments have given major insights into the topic. The readings obtained in the experiments was used to study the relationship between the distance and the width. These experiments were conduct for infinite number of goals. It is then confirmed that a relationship exists in three types of tunnels and that too with a correlation of more than 0.96.