Computer Animation Has Perceptual Limitations
Marc Green
The power of images is summarized by Confusious's saying, "One
picture is worth a thousand words." It is therefore hardly
surprising that photographs and other pictorial devices have long
been presented as evidence in court. More recently, computer animations
have become increasingly common in court. Presumably, they help
litigators visualize important events in an accident.
Computer animations fall into two distinct evidential classes:
- Demonstrative evidence: The animation is created by hand,
perhaps by an artist or draftsman. In this role, the animation
merely supplements witness testimony by illustrating key events.
For example, spatial relationships are difficult to convey in
words, so an animation may be useful to show events surrounding
an accident, e. g., the position, speed and trajectory of vehicles
and pedestrians.
- Substantive evidence: The animation is created by a
computer, which creates images based on input data and some algorithm
programmed into software. The event, such as an accident, is simulated
inside the computer.
When presented as substantive evidence, the animation may be presented
as reconstruction or visualization of the actual event. The idea is to show what
a witness or defendant could and could not have seen. For example,
it may claim to show what a driver or pilot saw before a crash.
The evidence may used to show that the pilot could not possibly
have seen the high tension wire or that the driver should easily
have seen the pedestrian.
In this application, it is important to remember that animations
have severe perceptual limitations.
- The computer graphic is only a simulation of the actual environment. It is impossible to say whether the animation omitted or introduced important visual elements. Perception is not absolute but depends on context. Virtually all judgments depend on background, the other objects in the scene, etc.
- There is usually some distortion of depth. Visualizations use mathematical models to simulate reality. Camera lens models used to create visual scenes may alter perceived depth and distance. Moreover, humans frequently use information from the two eyes (stereoscopic viewing) in depth perception. This information is lost in pictures and animations.
- Computer screens have limited ability to create images. Computer screens have much lower spatial and temporal resolution than real images. They may not be able to produce small objects or texture, which is a critical cue to depth and surface orientation. They are limited in their ability to render color and cannot create some subtle color variations. Computer screens also cannot produce very high contrasts, so object visibility may be underestimated. Lastly, computer monitors are not very bright and cannot produce light levels similar to a normal daytime. Since we see better in high light, this can result in a misleading estimate of conditions.
- The Small Scale Distorts. In court, graphics take a scene which may cover the viewer's entire visual field and reduce it in size to a few degrees of visual angle. Vision is clearest when an object is in the direct line of sight. The image then forms on the fovea, the retina's area of highest resolution. As images fall farther in the periphery of the visual field, they form on areas of lower photoreceptor density and are less visible. Graphics in court, however, are small so viewers don't use their peripheral vision. Attention usually moves with they fovea, so the attended object is in direct view. The judge and jurors know exactly where to look so they can point their eyes and focus attention to the critical part of the image. Lastly, the person involved in an accident may have had only a split second to detect the object while people in court can view the image for an extended period. These factors are almost certain to cause conspicuity overestimation. For example, there are companies that create forensic animations to show the driver's point-of-view of a pedestrian at times before an accident. In the image, the pedestrian appears only a few degrees from the line-of-sight instead of in the visual periphery, and the court, unlike the driver, knows that an accident is going to happen and so focuses attention and looks directly at the pedestrian. There is little or no background clutter to reduce conspicuity, no distraction from other cars, signs or attention to driving itself. Viewers in court are certain to think that they are seeing exactly what the driver saw and to overestimate the pedestrian's actual visibility.
- The court focuses attention. Elsewhere, I discuss how important attention is for perception. In the courtroom, the jury members know that they should carefully scrutinize the computer animation. In the actual event, however, the viewer may have had other tasks which absorbed some attention. Perception must always take attention into account, so the jury may be misled about the visibility conditions.
- The jury doesn't know what the viewer knew. Perception depends on experience, so a jury can never know or imagine what was perceived by the viewer at the time of the event. Much of our perception and behavior is guided by expectations that we develop as a result of our experience with the environment. The jury will not understand, for example, that the driver who used the same route everyday for a year and never saw a pedestrian was blinded by the expectation that there would be no one crossing the street. Moreover, the jury will not understand that this expectation-induced blindness is perfectly normal and adaptive human behavior and not gross misconduct.
Further, the animation may be subtly misleading and prejudicial even when
it is used to merely illustrate the accident events. Suppose an
animation provides an overhead view of a collision between vehicles.
Jurors may easily confuse their perception of speed and location
of objects seen in the animation with the actor's perceptions
of location and speed at the time of the accident. The jury may
draw incorrect inferences from the animation about the visibility
of vehicles, perceived speed, time to respond, etc. For example,
the animation may show a truck rear-ending a small car stopped on
a roadway. From the animation, it may appear that the truck driver
should easily have seen the car and braked in time. In fact, motion
perception is very poor for objects moving directly toward or away
from us, so the mistake in failing to discriminate a stopped
from a moving car directly ahead is a more a matter of innate human
capability than of gross negligence. Moreover, humans use size to
judge distance, so the fact that the stopped car was small would
further distort distance and speed perception. Lastly, the stopped
car in the middle of a roadway is contrary to driver expectations,
further delaying the truck driver's reaction. None of this would
be obvious from looking at the animation.
Conclusion
I have described some of the dangers inherent in misuse of animations as well as potentially misleading inferences which jurors might draw even when the animations are properly presented. Pictures are powerful because they are nonverbal and create an impression which cannot be easily diminished with logical argument - they appeal to the gut and not to the mind. Courts weighing admissibility of computer animations should therefore carefully consider their potentially misleading and prejudicial effects.
.
