Cognition and Perception

Transcription

1 Cognition and Perception 2/10/10 4:25 PM Scribe: Katy Ionis Today s Topics Visual processing in the brain Visual illusions Graphical perceptions vs. graphical cognition Preattentive features for design How can we use this knowledge to enhance visualizations? SIGGRAPH chapter: Ryan Orendorff Project pitch: Audrey Girouard Visual Processing Data Tasks -- Encode ---- Image --- View ---- Decode in the brain Visual Illusions: Example: Café Wall it looks like at the edges of the horizontal rows there is a tapering or shimmering effect but they are actually perfectly parallel Conflicting perceptual and cognitive cues Distortions in geometry (length, area, shape, angle) Distortions in intensity and color Understanding these illusions can tell us about our perceptual strengths, which in turn tells us how we should encode our data. Perspective illusions (mis)interpret visual elements as giving information about depth. Converging lines tell us that something farther away should be smaller so if it appears to be the same size as something in the foreground then we perceive it as larger. Note that converging lines used as axes would have a similar effect

2 Ponzo Illusion converging lines with two horizontal lines in between them. We judge the size of the horizontal lines based on the background. Get the impression that the line further away is longer. Muller-Lyer Illusion two lines of the same length, but one has angled pieces on the end facing out >-----< and the other has them facing in --. This causes them to appear different in length. Our perception of the neighbors affects our perception about the part of the visualization that we should be focusing on. Size Illusions Jastrow illusion lower shape appears to have a larger area because the top edge is being compared to the bottom edge of the above object. Titchener illusion an example of how we are better at judging lengths than areas and volumes Angular illusions like in the café wall, you may see tapering but the lines are parallel. We are not good at estimating exact angles of lines. We have a tendency to overestimate the size of acute angles and underestimate the size of obtuse angles Poggendorff illusion three line segments obscured by a rectangle, one on one size of the rectangle and two on the other. Due to our inability to estimate angles accurately we would guess incorrectly about which line is a continuation of which. Color and intensity illusions Checker-shadow illusion: the squares marked A and B on a checkerboard are the same shade of gray but perceived differently because your understanding of the scene brings in knowledge about shadows. We see color relative to the color of surrounding regions. This is an example of the effects of lateral inhibition. Hermann Grid see flashing dark spots at the intersections of white grid but they go away if you stare directly at them = 3 or more visual activation of negative space results: moiré patterns, vibration example: two black vertical bars but also see a third white bar in between

3 Mach Bands at the boundaries of two solid blocks of different gray hues it looks as if the edge closer to the dark band is lighter than the rest of the block. This is explained by the position of receptive fields in your eye and contrast between center and surrounding regions. Graphical perception Requires no conscious effort Preattentive Bottom-up information Graphical Cognition Requires conscious inspection Attentive Cognitive Top-down information Example: A field of numbers and asked to find how many 9 s are in there. Takes less time if you color the 9 s differently than the rest of the digits, because can quickly make the judgment using preattentive perception. When all digits are the same color the task requires more cognitive effort. Pop-out effect: For hue scattered dots are displayed and if one is a different color is it easy to find the target with preattentive processing For form one item is a different shape in a field of otherwise similarly shaped items; also easy to pick it out Eye movements Saccades rapid involuntary eye movements between fixations; not done consciously; allow you to get a macro view of the scene Fixations stop to look at a particular point Attend to feature outliers Measured through eye tracking Preattentive features

4 Include shape, color, texture, orientation, motion, groups, etc. Center-surround contrast with neighbors Separable feature channels considered independently Integral feature channels Implications for design Graphs can operate at the perceptual or cognitive level The most effective, or eye-catching graphs, operate (to some extent) at the perceptual level Must know which graphical attributes are easy to decode Empirical laws for perception Webers Law of just noticeable difference How big of a difference do we need to show in our markings for the viewer to be able to detect a difference? Line width example o Two lines with widths w and w+δw o Change of detecting a difference depends on the value of Δw/w o Just noticeable difference (jnd) Perceive relative differences o Differential threshold of a sensory stimulus o The ratio is more important than the magnitude o Appears to hold for many graphical encodings Exact jnd varies among people We perceive continuous variation as steps. For example, grayscale seen as discrete steps. Steven s Power Law for estimating magnitudes Encode a value x (magnitude of stimulus) Decode it to obtain a perceived value p(x) P(x) = kx α Constants k and α depend on the stimulus and encoding method There is a nonlinear relationship between x and p(x), which means we misestimate x. Perceive areas and volumes conservatively

5 underestimate large values relative to small ones overestimate small values relative to large ones perception affected by shape, so using a map area to encode data can lead to misinterpretation. For example Florida appears bigger than Georgia but they are different shapes so can t accurately compare. Estimating relative magnitude (Cleveland and McGill) extensive study of 7 graphical encodings for each chart (encoding) style, judge relative magnitudes 100+ subjects Which encodings have the strongest perceptual basis? ranking graphical encodings o position on a common scale Preferred o position along identical, non-aligned scales o length o angle, slope weak perceptual basis; main reason to avoid Avoid using to pie charts encode numerical o area values o volume o color Multivariate interactions multiple dimensions (variables), multiple encodings correlated or orthogonal dimensions use the strongest perceptual encoding for the most important variable redundancy gain dimensions reinforce one another interference loss: attention splintered difficult to view dimensions in isolation Example: Card game called Set Deck of 81 cards Each card has 4 features: color, symbol, number, shading

6 Goal is to find a set of three cars such that each feature is either the same on each card or is different on each card Conjunction search: Hue + form Searching for a single target with a combination of features Cannot identify the target preattentively Coffee graphic example: ingredient and amount of each is encoded. Ingredient is nominal, encoded by color and name. Amount is ratio encoded by a stacked barchart design.