Presentation on theme: "The Use And Misuse Of Machines That Simulate Human Cognition Hall Beck, PhD Office: 215 Smith-Wright Hall."— Presentation transcript:
The Use And Misuse Of Machines That Simulate Human Cognition Hall Beck, PhD Office: 215 Smith-Wright Hall
The Goals of Our Class 1) To learn about the appropriate and inappropriate uses of automation. 2) To speculate on how automation will impact our world and how we might study that world before it exists. 3) To provide you with an example of the scientific pursuit of knowledge. 4) To discover something about yourself and how you interact with your world
What Is Automation ? Any sensing, detection, information-processing, decision-making, or control action that could be performed by humans but is actually performed by a machine” (Moray, Inagaki, & Itoh, 2000) Automation is usually viewed as a continuum, ranging from manual control to full automation.
Some Quotes About Technology But lo!! Men have become the tools of their tools. - Henry David Thoreau
Some Quotes About Technology It has become appallingly obvious that our technology has exceeded our humanity. - Albert Einstein
Some Quotes About Technology We live in a time when automation is ushering in a second industrial revolution. - Adlai E. Stephenson
Some Quotes About Technology The first rule of a technology used in a business is that automation applied to an efficient operation will magnify efficiency. The second is that automation applied to an inefficient operation will magnify inefficiency. - Bill Gates
Four Generations of Artificial Environments (AEs) Where we have been, where we are, and where we are going
First Generation Unidirectional Communication-Information moves from the machine to the person but not the person to the machine.
Second Generation Bidirectional Communication-Information moves from the machine to the person and from the person to the machine.
Third Generation Virtual Reality-Information moves from the machine to the person and from the person to the machine. Ideally, the synthetic environment is indistinguishable from the actual environment.
Fourth Generation Life Simulation-The synthetic and actual environments are indistinguishable and the person does not know whether they are in an actual or synthetic world.
Automation Usage Decisions (AUDs) AUDs: Choices in which a human operator has the option of using manual control or one or more levels of automation (LOAs) to perform a task.
Some AUDs Are Commonplace Checkbooks may be balanced with a calculator or by mental computation Automobiles can be set to cruise control or the driver may operate the accelerator pedal Stock purchases may be based on the output of software programs or investors may depend upon their subjective assessment of the market
Some AUDs Have Historic Consequences Casey Jones Pearl Harbor Three Mile Island
Some AUDs Have Historic Consequences USS Greenville 2000 Election
Types of Automation Static: Level of automation is set a the design stage Adaptive: Level of automation varies depending upon the situation
Optimal And Suboptimal AUDs If it is assumed that the objective is to perform a task, the optimal AUD is to employ the level of control, manual through full automation, that maximizes the likelihood of a successful outcome. A suboptimal AUD is a choice to use a level of control that does not maximize the likelihood of successfully performing a task.
Types of Suboptimal AUDs Misuse is over reliance, employing automation when manual control or a relatively low LOA has a greater likelihood of success Disuse is the under utilization of automation, manually performing a task that could best be done by a machine or a higher LOA.
Errors Resulting in Misuse and/or Disuse Recognition Errors-Operator fails to recognize that an alternative, either automated or manual, is available. Appraisal Errors-Operator inaccurately estimates the utilities of the options. Intent Errors (also called action errors)- Operator knowingly selects the alternative that does not maximize the likelihood of task success.
Two Images of an Operator An operator is a single minded individual whose sole object is to maximize task performance An operator‘s decision to rely on automation is based on a number of contingencies only one of which is to achieve a successful performance.
Intent Errors and Decision Aids: Doing It Your Way When Your Way Is Obviously Wrong
Decision Aids And Intent Errors Probably no area of automation has proved more problematical than the introduction of decision aids Beck, Dzindolet and Pierce contended that much of the disuse of decision aids is due to intent errors That is, operators refuse “advice” from a decision aid that they know would improve their performance
200 “Training” Trials Participants viewed a series of slides on the computer screen, half of which contained a soldier in camouflage. Machine Absent: Pressed a “button” to indicate if the soldier was present or absent Machine Present: 1) Pressed a “button” to indicate if the soldier was present or absent and 2) Received the decision aid’s response
100 “Test” Trials Participants viewed a series of slides on the computer screen, half of which contained a soldier in camouflage. Machine Absent: Pressed a “button” to indicate if the soldier was present or absent Machine Present: 1) Received the decision aid’s “recommendation” and 2) Pressed a “button” to indicate if the soldier was present or absent
Rely On A Decision Aid: I Would Rather Lower My Grade Hall Beck, PhD Appalachian State University Mary T. Dzindolet Cameron University Linda G. Pierce Aberdeen Proving Ground
Objectives of Study 1 To determine the extent that persons would deviate from an optimal (rational) decision making strategy. To discover if automation misuse or automation disuse is the greater problem on this task.
Design 3 (Relative Performance: Inferior, Equal, Superior) x 2 (Feedback: Yes, No) between- subjects. Choice to base credit on self or computer as the dependent variable
Trial by Trail Procedure Present photo for.75 seconds. Participant responds via mouse indicating if the target was in the photo. Contrast detector then scans photo for human form. It attempts to determine if the target is in the photo.
Still More Procedure After 200 trials, participants in feedback condition are told how many errors they and the machine made. Participants in the No Feedback condition do not receive this information. Participants are either inferior, equal or superior to the contrast detector.
High Noon All participants are told 10 more trials will be conducted which will determine the extra credit that they receive. At this point, the participant must choose to base extra credit on self or machine (AUD).
Objectives of Study 2 To determine if feedback mitigates the bias against using automation. To discover what type of feedback is most effective in reducing this bias.
Forms of Feedback Trial by Trial: After each trial participant told if target was in the preceding photo. Cumulative: After 200 trials, participant told total errors made by self and contrast detector. Prior Results: After 200 trials, participants informed that persons who base extra credit on detector usually obtain more points.
Procedure and Design Same basic task as first two studies. Detector superior to all participants. 2 (Trial by Trial: Yes, No) x 2 (Cumulative: Yes, No) x 2 (Prior Results: Yes, No) between- subjects. Ten trials randomly selected. Choice to base credit on self or contrast detector.