Areas Home Areas Home Results Home Results Home Organization Home Organization Home Resources Home Graphics Computer Science Brown University Lab Home Visualization Research Group Update Search

David Laidlaw's advising pointers

1. How do I find a Research Project?
2. How do I print oversize pages?
3. How do I prepare a presentation of my work? (from the Brown Rhetoric Fellows program)
4. How do I prepare a presentation of my work? (notes from E. Tufte's course)
5. Are there any English courses for International Grad Students at Brown?

1. How do I find a Research Project?

Scientific visualization and computation research opportunities are always available! The process I suggest following involves first producing a proposal for a project that you'd like to do. Get buy-in from me and the group. Then do the work.

Producing the proposal is the tricky part, of course. Here's what goes into a proposal and some thoughts on how to create one:

  1. Your personal goals for a project. This section is, arguably, the most important. Without a good idea of your goals, it will be nearly impossible to find a satisfying project. Initially, your goals will probably be unrelated to a specific research project. They might include things like "find out what David is like to work with," "put my classroom math skills to work on a research project," "figure out if ____ research is something I want to do," or "combine math and painting somehow." As you develop the proposal, these goals will refine and likely become more specific. You'll also add some more research-related ones.
  2. A research topic or area to work on. In general, this needs to mesh with my own interests, or I won't be able to be an effective advisor on it. The scientific visualization web pages (http://vis.cs.brown.edu) are the best way to learn more about what the group is working on. Under results, look through the last couple of years of publications. There are also project pages that will give you a sense of how the different pieces of work fit together. My class web pages (http://www.cs.brown.edu/courses/cs237), particularly the Fall 2005 and 2007 instantiations, lists readings and project ideas that may stimulate your creativity. It also includes pointers to a number of research grants and contracts. The proposals themselves tell you what I plan to be doing. Based on these sources of information, make a list of some things that you find interesting.

My group meets weekly for a research meeting where one of the group presents some aspect of their work. Come to the meetings and attach yourself to an ongoing project. E-mail me to get onto the mailing list for weekly reminders. With some thoughts on your goals and some concrete specific ideas for projects, come talk to me. This is an ideal time to interactively converge on something we're both excited about. E-mail me with some times you are available and we'll set up a time to meet.

3) A work plan that describes the things that you will do to address your goals and the research problem/area.

4) A schedule, with concrete, quantifiable milestones every 1-2 weeks. Such a milestone might be "finish first draft of the related work section of final report." A schedule item like "read some papers" is not a milestone, is not concrete, and is not quantifiable.

5) Evaluation criteria. How will you define success? This should include not only evaluation of the research itself, but also of how well your goals were met. This evaluation will be the basis for a grade, if appropriate, so think of it as the "your grade will be based on ..." part of a syllabus. At the end of the project we will look back together at the project and determine what worked and what didn't in the context of these criteria and your goals.

The proposal is usually 2-4 pages. I prefer short and pithy, thanks! Expect to iterate on it 2-4 times before everyone is happy with it. Proposals that have not had their first iteration complete several weeks before they need to be finished are almost never successful. So start at least 3-4 weeks before it's due. If you want to do an independent study or reading and research project, the proposal must be completed BEFORE CLASSES START. No proposal, no class.

There are several examples of proposals (http://www.cs.brown.edu/~dhl/advising) off my home page. One of the best is Stephanie Lee's honors proposal and self-evaluation (http://vis.cs.brown.edu/docs/pdf/bib/Lee-2003-ISS.pdf).

Top

2. How do I print oversize pages?

You can use adobe acrobat 5.x (full ed.) to print out oversize pages (ie. posters:)) by tiling on smaller pages. The program works also very good for scaling up small documents, which might help you create quick posters. You can do all these directly on .pdf, .ps and .eps files without conversion.
CIS has a multi-user license for acrobat 5.x, you can use it in the public clusters. Alternatively you can download the software from the CIS web page and install on machine which is in the brown domain.

Top

3. How do I prepare a presentation of my work? (from the Brown Rhetoric Fellows program)

Look at http://vis.cs.brown.edu/docs/pdf/Brown-2003-RFH.pdf

Top

4. How do I prepare a presentation of my work? (notes from E. Tufte's course)

Presenting Data and Information
Edward Tufte
March 12, 2003

Topics


Introduction: Escaping Flatland
Fundamental Rules of Good Displays
The Smallest Effective Design
Reporting on Financial Information
Web and Screen Design
PowerPoint
Three Things to Demand of your Presenters When Making Important Business Decisions Making Better Presentations

Introduction: Escaping Flatland


When you create a display of quantitative information, you are trying to show multivariate (many- dimensional) information in a two-dimensional space. You are always trying to find ways of escaping Flatland.

Always be aware of the amount of resolution in your display. How much information (or, literally, how many numbers) are you able to convey with one single display? How does the medium you are using contribute to or detract from that resolution?

Fundamental Rules of Good Displays


(Example: Minard's map of Napoleon's march, The Visual Display of Quantitative Information, p. 41)
  1. Show visual comparisons. Your display should always answer your viewer's question, "compared to what?" For example, the Minard map shows comparisons of soldier deaths over time, in geographic space, and at different temperatures.
  2. Show causality. Give an explanation. Explain the dynamics and mechanisms for why the data is doing what it is doing. The Minard map connects the deaths of the soldiers with the temperature as they marched. Thinking causally can give your visual displays powerful effect and meaning. (Example: John Snow's cholera analysis, The Visual Display of Quantitative Information, p. 24)
  3. Show more than one or two variables. Show multivariate data. The Minard map shows time, latitude, longitude, direction, temperature, geographic features.
  4. Completely integrate your text and your graphics. Don't use legends! Make people do as little decoding as possible. And if you have a particular point you want to highlight, explain it and label it right on the display.
           Never use the "dreaded letter code", where something in your image is labeled with an "A" and
           the reader has to go to a separate table to find out what "A" is. Put your labels on the image
           itself--for example, if you have a time series with many lines, physically label each line with what
           it is; don't make people decode it with a legend. Give people less work to do in interpreting your
           information and they are more likely to read and understand it.
                      
     5.  Document everything. Where does the data come from? Who did this? Bureaucracies don't do
          things, people do things. Cite the author, cite the data source. This adds to your credibility, and it
          makes you more responsible for what you're putting on the page.
      
     6.  Content is KEY. Perhaps the most important point of all. Your presentation stands or falls on the
          quality, relevance, and integrity of your content. Often, the only goal of a good design is simply
          not to screw up your excellent content. And, likewise, the best design can't rescue bad or boring
          content. Minard's map has such power because of its content, and its purpose; it was meant by its
          author as an anti-war poster.
 
     7.  Show information adjacent in space, not stacked in time. Show your information physically on
          the same page so people can go back and forth and compare. If your information is presented in
          several graphs on separate pages, you just have one damn thing after another. PowerPoint
          presentations contribute to this--they encourage "relentless sequentionality".
 
     8.  Use small multiples. You may think it is easier to read one graph at a time, but in fact what might
          give more meaning is a number of small multiples of similar images on the same page. You lose
          some detail, but you gain the ability to compare. It is easier on your viewers--once they learn how
          to read one, it is easy for them to read the rest. It also gives you an inherent credibility, because it
          looks like you're presenting everything you have. It reduces suspicion that you're selecting your
          evidence. (Example: Galileo's sunspot engravings, Envisioning Information, pp. 18-19)
 
                     [Side Point: if you have time series graphs, try to scale the graph so that the average
                     absolute value of your slopes is around 45°. Exaggerated vertical scales to emphasize one
                     point can actually reduce the amount of information you see. (Example: Sunspot graph,
                     Visual Explanations, p. 25)]

The Smallest Effective Design


This is the theory that any move you make in your design should be as small as possible to avoid corrupting the content. There is no need to scream at your audience.

(Example: ear diagram, Visual Explanations, pp. 73-74) This diagram shows a finely-drawn diagram of the ear, and then super-thick pointer lines pointing towards the different parts of the ear. These thick lines distract from the content, making you think the drawing is about pointers rather than about the ear.

For another example, on a corporate organizational chart, people often put boxes around the various positions in the company. You really don't need those boxes; they can often be implicit. The staff names or titles and the lines connecting them are enough.

Always defend your content against the incursions of design!

Your design should come from the content of your display. The causality, the analysis, the multivariate data, your explanations, your reasoning--all these drive your design. The purpose of design is to assist thinking and allow people to have insight into your content. Your first question should be: what is the thinking task this display is supposed to help with? It is not about ease of use, or beauty.

The higher the resolution of your display, the better your reader's insights will be. A high-resolution display evokes responses for multiple constituencies--everyone can find something to look at. Something with a great deal of interesting, content-laden data, expressed in a clear way, can make people be so interested in the content of your display that they stop paying attention to your verbal explanation. (Example: Singer map, Visual Explanations, pp. 90-91)

Reporting on Financial Information


The most prominent parts of a spreadsheet are, of course, the top and bottom.

A great example of an actually useful spreadsheet: Gotti evidence spreadsheet, Envisioning Information, pp. 30-31.

Half of all financial analysis is an assessment of change. In order to show good comparisons over time, you do not always have to show the zero point; you do, however, have to provide whatever will give the full context of the change.

The key is to show the data and allow people to make relevant comparisons. Zero may be irrelevant. However, not showing enough context may skew your conclusions. A classic mistake in policy studies is to look too narrowly at a change over time, ignoring much of the history; it is easy to say that a policy had a particular effect when in fact the data may just be regressing towards the mean. (Example: Connecticut traffic deaths, The Visual Display of Quantitative Information, p. 74)

One good way to report on financial information is through micro/macro design. There is an entire chapter on this in Envisioning Information.

The idea behind micro/macro design is to have a display that shows the executive summary level view, as well as giving enough detail for credibility, all without being uncluttered. (Example: NY Times annual temperature graphic, The Visual Display of Quantitative Information, p. 30)

This is especially good in presentations, because it can speak to multiple constituencies. Some people may want to be looking at the macro view (e.g. it's warmer in July than in December), while others want to look at the detail (e.g. which individual days had the greatest highs and lows).

Ease of reading this graphic has nothing to do with the quantity of the information. Sometimes more numbers add clarity. And sometimes fewer numbers detract from clarity. One pie chart with only four numbers on it, and many colors, in 3-D, can be extremely off-putting to your audience. And you should never, never make business decisions based on a display like this. It says next to nothing, and distracts from the content with its presentation style.

Principles to follow when reporting on financial information:

  1. Standardize your data before showing it to anyone else. In particular, adjust for inflation. You must adjust for inflation whenever showing money over time, over more than two years.
                     [Note: annual reports of corporations are not required to adjust revenue for inflation. Not
                     adjusting for inflation may account for up to 25% of corporate growth as shown in annual
                     reports. Inflation does not equal productivity.]
                      
          Also adjust for seasonality in your revenue, where appropriate. Make sure you know where the
          highs and lows in your revenue stream really are, and how that should affect your interpretation of
          current data.
 
     2.  Document every display. Don't trust any financial report that doesn't have a footnote saying
          where the data came from.
 
     3.  Annotate your figures and your graphics. Explain everything.
 
     4.  Find excellent conventional designs and follow them. If something is widely-used and
          familiar--like bank statements, or the New York Times--it is probably a good model to use for
          your displays. Talent imitates, genius steals. Find four or five excellent templates, and use them.
 
          One good example is the "sparkline" graphs that come after the mutual fund information in the
          Wall Street Journal. This is a tiny graphic with intense design that shows the daily price of the
          fund over the past year.

          A person used to looking at these sparklines can glance down the page of the Journal and see
          funds with similar patterns through pattern recognition, to see which ones may use copycat fund
          structures; they can see which ones are generally up and which ones are generally down.
           
          It proves that human beings are able to process hundreds of thousands of numbers at once if they
          are presented in a clear, informative, contextual, familiar way. We don't need tinkertoy
          PowerPoint graphics all the time! We have a huge capacity for digesting information, and
          PowerPoint graphics trivialize our thinking. More on that later.

Web and Screen Design


(Example: National Gallery of Art's information kiosk, Visual Explanations, pp. 146-147)

Do not do web design by committee. Most screen design looks like it was BOGSAT design--Bunch Of Guys Sitting Around a Table.

Do not use a clever metaphor. Often your metaphor ends up being the bureaucracy responsible for the display. For example, news sites often have boxes around the different news departments (domestic, international, entertainment, etc.) as if to ward off the evil spirits from the other departments. Or if a computer programmer designs your search engine, you often end up with awful binary searches instead of something more human.

Put as much of your content up front as possible. Show the user all the things they can learn by exploring your site. Putting 200 to 300 links on your opening screen is not a bad thing, if they are cleverly designed. People will come back to a site they can get content from very easily, where they don't have a "long tunnel of entry". You don't want a site where the average download time exceeds the average time of the visit.

Use 90% of your screen for content. On a computer screen--a very low-resolution display medium--real estate is at a premium. Don't use up 60% of it on your navigation system or on marketing to your user. Beware of overly-elaborate graphics, "icons on steroids". On every screen, physically measure the area of the screen devoted to content. If it is less than 90%, go back to your designer and demand more.

No matter how good your screen design is, there should always be less of it. Good screen design is endlessly self-effacing and transparent to the user. Often all the design you need on your pages is a navigation bar at the top or side.

Great design comes from great designers. It does not come from user testing of your site. About all user testing will do is remove the disasters in your design.

PowerPoint


(Example: Boeing's slide explaining the cause of the Columbia disaster)

PowerPoint has great big serious problems. Mainly, it creates foreshortening of thought and makes us stupid. Here is why.

  1. PowerPoint slides have the lowest resolution of any display medium, even less than overhead projectors. It allows you to present only little squibs of information.
     Slides often use 60% or more of their surface area for branding, graphics, and other chartjunk. Usually
     40% or less of the slide is used for content. And because you have to use huge fonts, many true
     statements are too long to be displayed on a slide. Complex explanations are too long. If you are really
     trying to make a decision based on the presentation, abbreviating the truth for the slide is not good.
     The compressed language you have to use for PowerPoint is ambiguous or (sometimes dangerously)
     abbreviated.
      
     This is true for graphics as well. The average amount of numbers in a PowerPoint graphic is twelve.
     Printed reports can include thousands of numbers. Even graphics in Time magazine can include 50 or
     more numbers.

2. It is a presenter-oriented medium, not a content- or audience-oriented medium. It is very convenient

for presenters, but at a big cost to content and interest to the audience.

3. It gives you little space for comparison and causal thinking, instead encouraging you to present

slides one after another, stacked in time.

     Bulleted lists encourage generic, sloppy thinking. They encourage generic, meaningless phrases such
     as "Accelerate corporate profits". They don't encourage you to ask the who? what? why? how? that are
     really necessary for business decisions.
      
     "Nesting" is antithetical to how we think--it is an authoritarian structure that you have to jam your
     often complex content into. But nesting is, coincidentally, very parallel to computer programming. In
     fact, the metaphor for PowerPoint is that of a software house--hierarchical, nested, short lines of code,
     and tedious. And it is also marketing-oriented, advocacy- and not analysis-driven, fast-paced and
     commercial.
      
     This is not the right metaphor for teaching or analysis!

4. It puts the focus on the format of your presentation, not on your content.

5. It has the attitude (to the audience) of commercial marketing.

The key, again, is to use a high-resolution, effective presentation medium (paper) for showing data. And use a good analysis medium (talking) to analyze that data with the group. Don't expect to convey anything of content to people by jamming your data and explanation into an inappropriate medium.

Three Things to Demand of your Presenters When Making Important Business Decisions


(Example: Lessons from the Challenger disaster, Visual Explanations, pp. 38-52)

Show me the causality. Tell me what will happen if we do this. And what is the history, the mechanism, the theory, the link that you are using to predict that? Why should I believe you?

Show all relevant data. This can be problematic, because you may have tons and tons of data. But that is what 6-point type is for. Don't select your evidence.

Show me what I really need to know. If you are sitting in a bad presentation that you feel is not telling you what you need to know, get out of the room. Get out of the groupthink. Walk up and down the hall. Ask yourself: what do I really need to know to make this decision? What are the connections I need to make? And make your presenters answer that question.

Making Better Presentations


Show up early. You can head off serious problems (like the room being booked). It is a good time to say hello, ask names, find out about your audience, and advance your cause as people trickle in. It also shows respect for your audience.

Tell what the problem is, its relevance, and your solutions right up front. This gives people an advantage when trying to interpret your data. They'll ask better questions of you and the conversation will go farther if they have a preview of what you're trying to argue.

Never apologize for anything. Don't tell people you had the stomach flu last night, you have a sore throat, anything personal. Go as long as possible without using the first person. Focus on the content, not yourself.

Go to the particular, then the general, then the particular again. Get into some of the specifics of the data as fast as you can, and then before you've exhausted your audience, step back and contextualize, and then focus on a different particular that illustrates your general point.

Give everyone in the audience at least one piece of paper. This has symbolic value--you're leaving a record that you were there. It shows that you want your comments to have permanence. Also, it is the world's best method of information transfer. It shows much more information than a PowerPoint slide can, or than you can say over the course of your presentation.

We can do better than PowerPoint. Especially never use the PowerPoint templates. They are chock full of awful chartjunk. Projected art can be useful, of course--especially if you're leading a class on art history, for example--but it is much, much better to give copies of everything. And if you are using slides or an overhead projector, never, never, never do the "slow reveal" or "striptease" method of presenting.

Audiences deserve our respect. Don't patronize them.

Be careful with humor. It makes things memorable, it can let you use hyperbole and exaggeration to make a point. But it can also be offensive and embarrassing.

Don't use masculine pronouns as universals. Just go into the plural. For the past 240 years, at least, it has been completely acceptable to use the plural (they, their, etc.) as a generic pronoun. It doesn't cause a ripple; doesn't cause your audience to focus on your political correctness rather than your content.

Be enthusiastic. Don't hide behind the podium, get close to people where they can see you. Show that you believe in your own stuff and you're excited about it.

Finish early. Your audience will love you--and they will have a favorable memory of your presentation as well.

If you don't have good content, don't have the meeting.

Top

5. Are there any English courses for International Grad Students at Brown?

The Center for Language Studies and Office of Summer Studies are pleased to announce two new summer programs designed to assist international graduate students in advancing their English language skills.

  1. Intensive English Language and Acculturation Seminar for New International Graduate Students

    This pilot program for incoming international graduate students combines intensive English language instruction with an introduction to American academic culture and an orientation to life at Brown. The 10-day session will run from August 11th to the 22nd, involving approximately 12 pre-selected participants from three graduate departments. This summer's program is being supported by the Provost and the participating departments with cooperation from several offices on campus. In the future, we hope to expand the number of participating international graduate students.

  2. Intensive English Language Program for Continuing Graduate Students

    This Summer Studies program (developed with support from the Center for Language Studies, the Sheridan Center, the Graduate School, the Office of Career Services, the Writing Center and an ad hoc faculty advisory group) offers intensive English language course work designed specifically for graduate students. Students will improve their English language skills in:

    • Professional Presentations
    • Writing for Publication
    • Oral Communication

Please visit www.brown.edu/summer/iep/grad.html for course calendars, descriptions and enrollment information.

We are so pleased to contribute to these new projects and to share this news with you as the University makes strides to improve services for international graduate students.

Top
Page Owner: David Laidlaw Last Modified: Wed Mar 5 15:19:16 2008