Month: January 2010
the value of movement in personal communication
Real-time personal visual communication devices have not been commercially successful. 3G mobile video calling seems to have been a flop. Real-time visual communication isn’t a feature understood in marketing mobile communication devices. Is real-time personal visual communication a dead-end in the communications industry?
Recent research highlights the value of conveying movement in personal communication. Researchers Douglas Cunningham and Christian Wallraven compared faces represented as vertices, connected vertices, and surface-modeled, connected vertices. Reducing the number of vertices tended to degrade recognition of facial expressions in static representations of all three model types. But if expressive movement was incorporated in the models, reducing the number of vertices more than a hundred-fold (from 15,726 to 127) had little effect on expression recognition. Moreover, with movement represented, models with vertices-only performed about as well as surface-modeled, connected vertices. In addition, reducing the screen size of the representations from 512×384 pixels to 128×96 had little effect on expression recognition.[1] Simple, small, point-light displays that represent movement are efficient means for conveying facial expressions and emotions more generally.
Additional research shows that movement conveys information not present in static images. Actors performed realistic expressions of nine conversational emotions: agree, disagree, happy, sad, clueless (don’t know the answer), thinking, confused (don’t understand question), disgust, and pleasantly surprised. Cunningham and Wallraven tested persons’ ability to recognize (in a 10-alternative, non-forced-choice task) recordings of these expressions: a video sequence running from neutral expression to peak expression, and a static image at peak expression. The over-all accuracy of recognition was significantly higher for the video sequence than for the static image (78% versus 52% correct identification). [2]
A variety of additional experiments further identified the value of movement. Cunningham and Wallraven reduced the video sequence to the last 16 frames and expanded the static presentation to those sixteen frames temporally sequenced in an over-all image grid. Participants significantly more accurately recognized facial expressions in the dynamic 16-frame presentation than in the static 16-frame grid (roughly 75% to 60%). Scrambling the order of the frames in video presentations roughly eliminated the advantage of video. Playing the video backwards significantly lessened the accuracy of expression recognition. At least 100 milliseconds of temporarly integrated, forward-sequenced images seems necessary to capture the value of movement information for recognizing facial expressions.[3]
Video news channels illustrate the communicative power of facial movement. Try watching a major, anchor-based news broadcast with the sound turned off. The facial expressions of news anchors are quite extraordinary. They intently focus their eyes straight out at the viewer, exaggerate head movements and facial gestures, sharply punctuate their facial gestures with pauses and rapid dynamics, and expressively communicate concern and urgency as they report a father discovering a man hiding in the bushes and looking into his daughter’s bedroom. The news anchors’ facial expressions powerfully attract viewers’ attention and shape viewers’ emotional responses.
The rapid take-up of iPhones demonstrates that good device design can transform a broad product space. Smart phones, e-readers, and various other electronic devices are proliferating. At least one informed industry participant sees a bright future for see-what-I-see communications devices. Developing a real-time visual communications device will require considerable innovation in device design. But with a good biological and ecological design, such a device could have a major impact on the communications industry.
Notes:
[1] See Cunningham, Douglas W. and Christian Wallraven. The interaction between motion and form in expression recognition. Proceedings of the 6th Symposium on Applied Perception in Graphics and Visualization (APGV 2009), 41-44. (Eds.) Mania, K., B. E. Riecke, S. N. Spencer, B. Bodenheimer, C. O‘Sullivan, ACM Press, New York, NY, USA (2009). These findings don’t imply that video quality matters little. Rather, they point to the importance of a biologically and ecologically informed analysis of video quality. A point-light display provides a high-fidelity representation of a small number of points of motion.
[2] See Cunningham, D. W. & Wallraven, C. (2009). Dynamic information for the recognition of conversational expressions. Journal of Vision, 9(13):7, 1-17, http://journalofvision.org/9/13/7/, doi:10.1167/9.13.7. The advantage of video depends significantly on the specific conversational expression. Video is hugely advantageous for communicating agree and disagree. That’s probably because slight head movements vertically and horizontally, respectively, tend to characterize these expressions. Happiness, in contrast, was the only expression correctly identified significantly more accurately with a static presentation than with a video presentation. Compared to a static expression of happiness, a video expression of happiness may be more extensively interpreted, e.g. as actually indicating deception or manipulation. The experimental determination of correct interpretation doesn’t control for different levels of interpretation.
[3] Id.
natural gas is the answer
Benjamin Franklin was a founding father of the United States. In a letter to a science academy in 1781, he proposed that the academy seek an invention that would “render the natural Discharges of Wind from our Bodies, not only inoffensive, but agreable as Perfumes.” Franklin declared:
Are there twenty Men in Europe at this Day, the happier, or even the easier, for any Knowledge they have pick’d out of Aristotle? What Comfort can the Vortices of Descartes give to a Man who has Whirlwinds in his Bowels! The Knowledge of Newton’s mutual Attraction of the Particles of Matter, can it afford Ease to him who is rack’d by their mutual Repulsion, and the cruel Distensions it occasions? The Pleasure arising to a few Philosophers, from seeing, a few Times in their Life, the Threads of Light untwisted, and separated by the Newtonian Prism into seven Colours, can it be compared with the Ease and Comfort every Man living might feel seven times a Day, by discharging freely the Wind from his Bowels? … surely such a Liberty of Expressing one’s Scent-iments, and pleasing one another, is of infinitely more Importance to human Happiness than that Liberty of the Press, or of abusing one another, which the English are so ready to fight & die for.
The Internet has returned freedom of the press to the practical ideal of Franklin’s time. Even better, natural gas is now more valuable than perfume. For a sweet-smelling future, natural gas is the answer.
describing name frequency distributions
The most popular given names have become much less popular over the past two centuries. In the U.K. about 1800, about 85% of males and 82% of females had names that were among the ten most popular given names for males and females, respectively. Apart from temporary effects of the Norman Conquest in 1066 and the Black Death in the mid-fourteenth century, given names seem to have had a similar distribution from 1000 to 1800. But after 1800, the name distribution flattened. By 1994, the share of males and females with given names among the ten most popular given names had fallen to 28% and 24%.[1] That seems to me to be a quite astonishing change in an important class of symbols.
Describing the name distribution as flattening is a simple way to describe the change from 1800 to the present. More specifically, plot name popularity (frequency of a name divided by the total number of named persons in the sample) by popularity rank. Approximate this plot across some range of ranks by a line. The slope of that line has flattened over the past two centuries. That’s equivalent to the complementary distribution function of name frequency increasing in slope at the high end of the name frequency distribution.[2]
My work on given name frequency distributions does not support claiming that given names follow a power-law distribution. To the extent that I have in the past made such a claim, please recognize that I provided no statistical support for that claim. You can find evidence that I don’t take such a claim seriously. I’m interested in understanding major changes in symbolic choices such as the change in name popularity over the past two centuries. Given this interest, expending effort on estimating a stationary statistical model for the name distribution doesn’t seem to me worthwhile. I hereby explicitly renounce any claim that given names follow a power-law distribution.
Moreover, I heartily recommend recent work on estimating a power-law model, evaluating its goodness of fit, and comparing it to alternative statistical models. With their article, “Power-Law Distributions in Empirical Data,” Aaron Clauset, Cosma Rohilla Shalizi, and M. E. J. Newman provide a clear exposition of power laws, useful estimation strategies (see especially equation 3.7), and analysis of twenty-four real-world data sets.[3] Even better, they have made available on the web code, in several languages, that implements their estimation methods. They have also made their test data sets available through the web to the extent that they could. In short, their work is an outstanding example of actual, significant advancement of public knowledge.
Notes:
[1] These figures are from Galbi, Douglas (2002), Long-Term Trends in Personal Given Name Frequencies in the UK, Table 1.
[2] Name frequencies divided by sample sizes give name popularities. Moving leftward from the high-end name frequencies and assuming distinct frequency ranks, the empirical complementary distribution function increases by constant probability increments equal to the inverse of the total number of names in the sample. Hence, to scaling parameters, the popularity/rank plot is equal to a transformation about the x-y axis of the complementary distribution function. Plotting in log-log space eliminates the effects of the scaling parameters.
[3] One of their data sets is surname frequency from the U.S. Census of 1990. Their analysis favors for these data, above a minimum frequency threshold, a power-law with exponential cut-off. A power-law distribution and a log-normal distribution are not clearly rejected. However, these statistics shouldn’t be taken too seriously. Compare the source Census surname data set to Clauset, Shalizi, and Newman’s constructed surname frequency data set. They constructed surname frequencies from surname frequency percent shares for surnames with shares above 0.005, reported with only one significant digit. Even at their estimated x-min (upper frequencies), the reported surname share has only two significant digits (0.0045). Hence, while the computed frequencies are reported with seven significant digits, their accuracy in most cases is much less. That the analysis of surname frequencies doesn’t employ good data probably isn’t important. A statistical model for surnames seems to me less interesting than a statistical model for given names. Moreover, as discussed above, I think the most interesting issue for given names is distribution dynamics. I hope that smart statisticians will work on understanding given-name distribution dynamics.