Excellent article Enrico - shared it with my class. I think the classical teaching method of precision of reading is a good guide, but we also need to take into account, as you say, effectiveness. The appropriate choice of encoding depends on not just the type of the data but also the semantics (great point about the part/whole argument in favour of pie charts). I also wonder what role convention has to play? If people expect some types of data to be presented using a given type of visualization, should we take that into account, even if it's not empirically ideal in a lab-based perception study? Maybe? Or maybe we should be pushing for adoption of new evidence-based designs.
I am very happy to hear this article resonated with you. I have been thinking about this problem for a while and I think it's crucial to make some progress. Yes, conventions may play a role and also culture! I am not sure how to take that into account. Your last question is a pertinent one. We have initial lab studies showing that basic visual features have some pretty strong affordances but whether these affordances have an impact on performance (and performance in doing what?) remains an open question. To me it seems this problem is big enough that many people should be working on it! Thanks for your message! I am super happy to hear you sent this to your students.
I like this post, and I think you are unto something. What difficulty I find with your scientific approach to validating the effectiveness of expressiveness is that both measures are inherently subjective. You are going to be measuring the quality of the visualizations based on perceived understanding (effectiveness) to determine whether the graph type (expressiveness) has any impact. It looks to be more of Sociology issue in that framing.
Interesting - with the New York student question would a cartogram not be better?
This would look nicer, add information about New York geography and go well with a colour or lightness / value.
Glancing at the ratio that was calculated looks fairly even - if large discrepancy likely age of cars / speed limit of roads would account for the difference,
Excellent article Enrico - shared it with my class. I think the classical teaching method of precision of reading is a good guide, but we also need to take into account, as you say, effectiveness. The appropriate choice of encoding depends on not just the type of the data but also the semantics (great point about the part/whole argument in favour of pie charts). I also wonder what role convention has to play? If people expect some types of data to be presented using a given type of visualization, should we take that into account, even if it's not empirically ideal in a lab-based perception study? Maybe? Or maybe we should be pushing for adoption of new evidence-based designs.
I am very happy to hear this article resonated with you. I have been thinking about this problem for a while and I think it's crucial to make some progress. Yes, conventions may play a role and also culture! I am not sure how to take that into account. Your last question is a pertinent one. We have initial lab studies showing that basic visual features have some pretty strong affordances but whether these affordances have an impact on performance (and performance in doing what?) remains an open question. To me it seems this problem is big enough that many people should be working on it! Thanks for your message! I am super happy to hear you sent this to your students.
I like this post, and I think you are unto something. What difficulty I find with your scientific approach to validating the effectiveness of expressiveness is that both measures are inherently subjective. You are going to be measuring the quality of the visualizations based on perceived understanding (effectiveness) to determine whether the graph type (expressiveness) has any impact. It looks to be more of Sociology issue in that framing.
Interesting - with the New York student question would a cartogram not be better?
This would look nicer, add information about New York geography and go well with a colour or lightness / value.
Glancing at the ratio that was calculated looks fairly even - if large discrepancy likely age of cars / speed limit of roads would account for the difference,
Chris
www.chrisweatherburn.com
You need to separate maps to do that. Which is fine, but comparing across regions is hard.