SIGGRAPH 2004 - Course: Seeing, Hearing, and Touching: Putting It All Together August 8, 2004 (Sunday) by the University of British Columbia HCI Faculty [These are just some rough notes I made, there were lots of cool visual examples though - I ended up applying to UBC for grad school because of these great lectures.] VISION SCIENCE Visual Primitives diameter/area summaries made rapidly some integral, some separable size is both length and width = act together, cannot just separate width hue and lightness in 2D color space Rapid Vision Visual Search: color, orientation (2D), length, curvature, motion 1) convexity - target/distractor surface convexity - bumps and depths top-bottom easier than left-right 2) 3D orientation - rectangularity vs. curvature 3) shadows Metacognitive Gap There only exists a few things in attention at a time you access when you need it present simultaneously but not represented similarly virtual rep. - world = external memory focus on object when need to for the task at hand no universal representation of an image 1) different people see same world differently - cultural flicker paradigm 2) navigation = NATURAL process - infospace demand-driven vision mechanisms already exist to complete our perception system "natural born cyborgs" (Clark 2003) to environment 3) 4-5 things can be held at a time (short-term capacity) Attention: pools info from 4-5 links to a single Nexus Search for presence of change I. 1 change at a time - "simultagnosia" other: audition - "change deafness" haptics - "change numbness" vision + audition, audition + haptics, vision + haptics II. Decomposition of objects - rendering is expensive what to render and what in detail flicker paradigm change blindness useful for control of attention III. Space Perception - layout, gist (layout - abstract arrangement of items in scene) Shaker Paradigm - compare performance translation - X rotation - 4 degrees ok scaling - X if jumps aren't too big, ok fisheye - search slows down because clutter but transition has no effect unconscious non-atten. symbols - gist, layout, etc. visuomotor, oculomotor response - corrective - eyes follow it manual pointing - hand knows even if change not registered What system? Atten, gist, slow rational decisions - only one studied, displays are thus made for conscious observers - what about for unconscious? How system? Eye movements, hand, head, arm - INNER ZOMBIE? for control of mouse, sixth sense, natural functions for alertness (like unconscious driving after long hours?) change felt (gut feeling) vs. change seen - 1/3 observers dif. sensitivities to change "Mindsight" Coercive Graphics? (advertising?) Unconsciously hijack system, automatic seeing, more natural control, eg. magicians control people's attention Perception - not on detailed picture - "just in time" - similar to operation of other senses Nonattnal Process - guide inner zombie Vision - result of several systems - integration key Navigation offload cognition - use visual system to import offloaded as we need I. External Rep: 57 x 48 = __6 mental buffer: [7 x 8 = 56] - synthetic? Help human perform some task (Hofstader 1979, Topic Graphics) eg. Layers 1. Cartography - creating visual layers, color => regions 2. Graphs - minimize crossing edges - but in layers ok - size, brightness, locations use perception to maintain edge crossing 3. Layers in Occlusion - extravision 3D - pretty but not useful? (Van Wijk + Van Selow, InfoVIS) N pairs of (value, time) N large find std. day patterns, how distribute over years, etc. => hierarchical clustering, binary hierarchy, dendrograms Linked highlighting - calendar line graphs task analysis may lead away from obvious choices instead clusters - meaningful derived space highlighting - linked with mult. pair coordinated views exploratory data visualizer real world navigation only partially understood Kevin Lynch - environmental cognition - synthetic landmarks motion beyond rotate, translate, zoom - speed-dependent automatic zooming multiscale navigation zoomable UI - Pad++ Space-scale diagrams - Furnas and Bederson Semantic Zooming instantiate new objects as there's room to draw them zoom and panning anisotropic cost - zooming vs. panning - cannot see all, optical flow, shortest (UIST) speed-dependent zooming - scale x speed = constant, fast = zoom out, slow = zoom in Hinkley - gets bigger as requested pan gets faster (very intuitive!), fisheye view Focus + Context - avoiding disorientation problem - maintaining orientation hard for lots of detail 2 windows - thumbnail view, single combined view distortion - nonlinear fisheye - pliable surfaces hyperbolic - focus and context stretch surface - transformation and magnification Multiple foci - polyfocal function - makes things in btw smaller hyperbolic geometry h3-3d space tree - interactive representation of data