Trust in Automation Muir and Moray

1
Trust in AutomationMuir and Moray

• HUMF/IE 8541 Decision Support Systems

2
Trust in Automation

• Before Neville Moray decided to study the issue
  of trust in automation in the early 1990s,
  people didnt think of trust and automation
  or computers in the same sentence.
• Moray and his graduate students, John Lee, Bonnie
  Muir, etc. started people thinking about the fact
  that people really did place trust (or distrust)
  in computers in much the same was as they form
  trust relationships with people.
• They were pioneers in computers and trust.

3
2 Experiments

• Context Milk processing plant,
• Task Supervisory control task, computer can be
  used to supervise processes in a factory,
• Operators have to decide when to intervene, when
  not to.
• Rating Used subjective ratings of operators
  trust in the automation,

4
Goals of Work

• To predict and optimize operators intervention
  behavior on the basis of their degree of trust
  in the automation,
• To identify what properties of a system determine
  users trust,

5
Results of Experiments

• Trust in automation reflected in perception of
  machines competence,
• Trust was significantly reduced by any sign of
  incompetence, even one that did not effect
  overall systems performance.
• Models of trust between humans capture some
  important aspects of human-machine trust.

6
Experimental Questions

• Can operators meaningfully rate their trust in a
  machine?
• How is trust related to the properties of the
  system?
• How does trust relate to intervention behavior?
• How does trust relate to monitoring behavior?
• How does distrust spread with respect to a
  system?
• Are models of human-human trust relevant for
  human-machine trust?

7
Experiments

• Experiment 1
• Semi-automation for pump
• Manual control of heater
• Experiment 2
• Semi-automation for pump
• Automatic control of heater

8
Experimental Setup

• Simulated milk processing plant
• Must heat raw milk to specified temperature, for
  a limited amount of time.
• Must kill bacteria with out scorching milk and
  ruining flavor,
• Operator must maximize milk output for the plant.

9
Questions

• What other factors do you think come into play in
  this situation?
• How can the interface be redesigned to help the
  user more?
• What questions should you ask (about users, task,
  context, etc.) in order to do this re-design?

10
Usability Tests

• Why were the questions stated in an open ended
  manner?
• How many tests is enough tests?
• How many users are enough users?
• What parameters/variables are important to test
  for a webpage?
• How many should/can you assess in any given test?

11
Usability tests on Webpage

• Test 1 1 user
• Test 2 2 users
• Test 3 6 users
• Test 4 1 user
• Test 5 ?

12
Webpage Usability Tests

• Parameters to be assessed
• Appeal
• Image,
• Colors,
• Layout,
• Understandability of text,
• Appropriateness and completeness of text for
  tasks
• What tasks do people need to perform with this
  webpage?

13
Trust in Websites

• People trust websites with pictures of people
  more than those that lack pictures of people.