Identifying Design Principles

From CS 294-10 Visualization Sp10

Jump to: navigation, search

Lecture on Mar 17, 2010

Slides

Contents

Readings

  • Pictorial and verbal tools for conveying routes, Lee & Tversky (pdf)
  • Rendering effective routemaps, Agrawala & Stolte (pdf)
  • Identification and validation of cognitive design principles for automated generation of assembly instructions, Heiser et al. (html)

Optional Readings

  • Designing effective step-by-step assembly instructions, Agrawala et al. (html)

Demonstrations

Jeffrey Patzer - Mar 17, 2010 11:30:28 am

I really like the route based map idea. This would be a great system to use on a mobile handset, considering the screen size layout. I think however, something that would greatly improve the effectiveness of map technique, would be to use landmarks and street names to denote a turn. A lot of times, road names can be hard to find when driving. Usually landmarks are easy to find and quicker to locate. I think it is common to hear verbal directions interspersed with landmarks and other features to help the person navigate, and having those sort of directions used in a route map could also be extremely useful.

RyanGreenberg - Mar 17, 2010 12:46:14 pm

I thought the Heiser paper made the process for codifying design principles seem very approachable. Steps one through three were about getting concrete data about what makes assembly instructions good, step four was about implementing that, and step five was testing it. The steps seem quite straightforward, though I wonder how much of this was because of the limited domain under consideration (instructions for assembly). How much harder does this process become when you consider much larger or more abstract domains like the ones we've discussed in class (e.g. appropriate use of color)?

One question: the computer generated instructions (figure 7) are in color including red arrows while the other two options (figure 8) are black-and-white line drawings. This seems like an unfair comparison. Why not compare the hand-drawn and factory instructions with black-and-white computer-generated instructions?

Stephen Chu - Mar 21, 2010 11:14:16 pm

Conveying Routes: I never thought too much about the parallel structure common to the descriptions and depictions of routes, even though looking at Google Maps shows a clear mapping between descriptions and depictions. In my experience, it helps to use midpoints, not just use start-end segments, when describing a route. It helps to know what I should be encountering/seeing along a path to make sure that I'm on the right path and to know how much more I have to travel. E.g. "From the corner of Regent and Parker, go down Parker to Telegraph and Parker. You should pass by a 7-11."

The paper states, "Although all participants used both rectangular and round landmarks, the rectangular ones were used as a default, and the round ones in special cases where the landmark was round." I've asked myself this question a few times before. Why do people generally use rectangles as the default over round shapes? Maybe not an important question, but it'd be interesting to know the reason.

Heiser et al. Paper: This was an interesting read. The paper listed design principles to be followed. I'd be interested in knowing the ranking of the importance of each of these principles in improving the construction efficiency.

Zev Winkelman - Mar 24, 2010 11:47:36 am

Tversky and Lee:

Limitations of depictions (1.1) - specificity ... limits their expressiveness ... it is difficult to convey abstract concepts such as justice and freedom or relations such (as) counterfactual and hypothetical pictorially.

I'm glad they included a reference for this (Stenning and Oberlander 1995), cause I'd like to see how the conclusion was reached.

Other than that I thought the most interesting part was the two simple inference rules (2.4):

-continuity - if a starting point is missing, it is the same as the previous end point or vice versa

-forward progression - when two reorientations occur successively, a forward movement is implied between those two reorientations

These two rules seem to allow the finding in (4.2) that verbal and graphic elements map onto one another.

Heiser et al:

Interesting application of methodology similar to Tversky and Lee.

The results aren't too surprising.

The most interesting claims are those about generalizability to other domains:

(Introduction)

-cognitive design principles identified ... can be generalized to ... other products

-approach ... can be generalized to other domains

(Conclusion)

-design principles ... can be applied to the design of visualizations in similar domains

I wonder whether this is true in domains where there is less convergence on what best practices, or clear instructions, are.

DavidZats - Mar 24, 2010 03:00:57 pm

Much of this lecture was focused on route generation. I really enjoyed this lecture, because it brought up some interesting aspects about visualization. Specifically, this lecture provided examples of how distorting or omitting information allows the viewer to achieve greater clarity. Properly distorting information such as the distance traveled on roads allows the viewer to more clearly discern the turns that must be made to reach the destination. Additionally, removing extraneous information helps the viewer to more clearly focus on the information necessary to complete the task. Through these examples, route-maps provide an excellent case study of how to abstract information so that it may be more clearly understood by the viewer.

One of the questions I had about this work was why current online route mapping software does not make use of the route-maps functionality advocated for in this lecture. Given how much it would aid the viewer in reaching the destination, I am surprised that this has not been readily adopted.

Boaz Avital - Mar 24, 2010 04:54:31 pm

The studies seem to have a common strategy of getting a number of human subjects to make their own versions of the goal products (maps, instructions) and then trying to emulate programmatically the best features of what they came up with. This has good and bad features. The good part of it is that people will create visualizations that they believe are easy for them and others to understand, and hopefully when emulated that will be better than just an approach not based in what people already do. The bad part could be that by relying on laymen and not experts, you could be missing out on simple, just beyond obvious, and extremely helpful improvements to your visualization.

Arpad Kovacs - Mar 29, 2010 02:20:59 pm

I found the assembly instructions paper to be very intriguing, since it takes a seemingly subjective question ("What are the attributes of effective assembly instructions?"), and provides a computerized solution based on the results of scientific, quantitative studies. It appears that high spatial-ability/experience participants, (and consequently the effective instructions they provided) were process-driven, and focused on the steps to put the TV stand together in action diagrams, unlike the less efficient results-driven structural and part diagrams used by low SA/E participants. Arrows and clear ordering in particular seem well-suited to conveying placement, direction, and action, while words tend to be distracting and should only be used sparingly, or for clarification in our own visualizations.

My only concern is that the hand-drawn and factory-provided instructions are black-and-white, while the automated system generates color diagrams. I think that the results would be more comparable if the computer-generated result were black-and-white as well, since as we saw in the perception lecture, something as seemingly innocuous as painting the arrows red could trigger preattentive processing that bias the results (especially completion-time).

LineDrive also shares this process-driven approach: rather than focusing on the source and destination as geographic locations like in traditional maps, it instead emphasizes the individual turns and roads that the driver needs to take. It would be interesting to apply this concept to other areas, such as mass-transit maps, or visualizing cooking recipes.

Jonathan Yen - Mar 29, 2010 03:39:40 pm

Again, I think the LineDrive project was a really well thought-out implementation of a map visualization.

The assembly instructions reading is pretty interesting. I wonder if there are studies that compare static instructions compared to having animated assembly instructions, as I think that an animated sequence of assembly instructions would help to further reduce the assembly time.

Jon Barron - Mar 29, 2010 10:25:49 pm

Tversky & Lee

I suppose this paper is somewhat useful, as a "simulation of a simulation", as they say. Still, a paper that says that maybe directions can be translated to maps and vice versa is much less impressive than an implementation of a system that does just that. Line Drive spoiled me, I guess.

Agrawala & Stolte

I like this paper, but I still can't shake the nagging feeling that the optimization techniques are lacking. I don't really see the need for simulated annealing, as the majority of the loss function seems differentiable (suggesting that gradient descent might work). The only non-differentiable parts I see are missing and false intersections, which don't seem like they'd be a problem in gradient-based optimization, as one could simply penalize both with infinite cost, and then assuming that the initial map has all true and no false intersections, optimization would preserve that structure. I also find it slightly unsettling that the final map seems very dependent on the initialization, which a convex formulation could address.

Akshay Kannan - Mar 30, 2010 01:33:42 pm

The LineDrive project is an extremely useful visual representation of a driving route. Because only the characteristics useful to the user are preserved in the visualization, the user can much more easily see the route as a whole, along with the important intersections. An interesting idea I saw here, as well in other map representations, is discarding the precise nature of the route in favor of a line-based representations, where turns in the route are simplified into a form more easily viewable. This seems like a very good idea for user simplicity and eliminating additional details.

Shimul Sachdeva - Mar 31, 2010 03:52:39 am

I really like the demos in class on the current topic. The grocery store aisle project and the project that maps streets of San Francisco were especially interesting. Is there an efficient way to automate the process of taking pictures? Perhaps a device that when mounted on a car just takes images as it drives by and generates images in real time. The project on assembly instructions (with the TV table experiment) is pretty interesting too. I wonder if this technology can be used with existing CAD software to automate the generation of assembly instructions?

Jaeyoung Choi - Mar 31, 2010 07:18:19 pm

I find LineDrive project to be especially interesting as I always struggled when I tried to use google map to plan (print or memorize) route before driving. It's interesting to recall Tufte's negative view on distorting visualization of data to convey the idea, when in this project, the route map is effectively omitted or distorted for better understanding of the whole route without losing detail. One suggestion is that it would be easier to find where to turn if the cross street that comes right before the turn is also drawn and labeled.

Prahalika Reddy - May 12, 2010 07:27:28 am

I really enjoyed this lecture. I especially liked all the images and examples that we looked at. One of the projects that I found most interesting was the one about fitting all the buildings on the street into one picture. The method I liked best was the graph cut. It showed all the buildings, but didn't distort any of them.

Another interesting problem was that of route visualization. I've never thought about how it could be more effective to be less accurate to be more effective. I think that's something that's not very obvious a lot of the time.

Priyanka Reddy - May 14, 2010 06:29:42 am

What I found most interesting in this lecture were all the images that combined two perspectives. First, I don't think most people, when they see those pictures, notice that there are two different perspectives combined and that that combination is a physical impossibility in the real world. But, what really makes it interesting is that those impossible combinations are used in order to make the picture look less distorted and more real.



[add comment]
Personal tools