Download presentation
Presentation is loading. Please wait.
1
Role of Effective Communication in Trust-Building: Application to Human-Computer Interaction
FA PI: Mohammad Maifi Hasan Khan Department of Computer Science and Engineering University of Connecticut Presented at the 2016 Trust and Influence Program Review Air Force Office of Scientific Research Arlington, VA, June 15, 2016
2
Team Members Emil Coman Mohammad Maifi Hasan Khan (PI) Ross Buck
Assistant Professor Department of Computer Science and Engineering University of Connecticut Emil Coman Research Associate Health Disparities Institute UConn Health Center Ross Buck Professor Department of Communication University of Connecticut
3
Typical Control Centers for Safety-critical Systems
4
Characteristics of Safety-critical Systems
System operators have to process a large volume of information in real-time. Systems may fail/malfunction for many different reasons - Environmental factors Communication error Hardware/software failures …………….. Such failures are likely to effect users’ emotions significantly which is not currently considered while designing human-computer interactions for such systems!
5
So, why should we care about users’ emotions?
6
Affect, Reason, and Involvement
7
EMOTION, DECISION MAKING, AND RISK
Emotions can both enhance and undermine effective decision making.
8
EMOTION, DECISION MAKING, AND RISK
Emotions can both enhance and undermine effective decision making. Decision literature (e.g., the Risk as Feeling hypothesis) typically defines emotions simply in term of positive and negative valence.
9
EMOTION, DECISION MAKING, AND RISK
Emotions can both enhance and undermine effective decision making. Decision literature (e.g., the Risk as Feeling hypothesis) typically defines emotions simply in term of positive and negative valence. Affective neuroscience has found evidence for specific neurochemical systems underlying a variety of emotions that may be involved in risk/decisions.
10
EMOTION, DECISION MAKING, AND RISK
Emotions can both enhance and undermine effective decision making. Decision literature (e.g., the Risk as Feeling hypothesis) typically defines emotions simply in term of positive and negative valence. Affective neuroscience has found evidence for specific neurochemical systems underlying a variety of emotions that may be involved in risk/decisions. POSITIVE INDIVIDUALIST (CONFIDENT, SECURE, SATISFIED)
11
EMOTION, DECISION MAKING, AND RISK
Emotions can both enhance and undermine effective decision making. Decision literature (e.g., the Risk as Feeling hypothesis) typically defines emotions simply in term of positive and negative valence. Affective neuroscience has found evidence for specific neurochemical systems underlying a variety of emotions that may be involved in risk/decisions. POSITIVE INDIVIDUALIST (CONFIDENT, SECURE, SATISFIED) NEGATIVE INDIVIDUALIST (NERVOUS, UNSURE, FEARFUL)
12
EMOTION, DECISION MAKING, AND RISK
Emotions can both enhance and undermine effective decision making. Decision literature (e.g., the Risk as Feeling hypothesis) typically defines emotions simply in term of positive and negative valence. Affective neuroscience has found evidence for specific neurochemical systems underlying a variety of emotions that may be involved in risk/decisions. POSITIVE INDIVIDUALIST (CONFIDENT, SECURE, SATISFIED) NEGATIVE INDIVIDUALIST (NERVOUS, UNSURE, FEARFUL) POSITIVE PROSOCIAL (PROUD, TRUSTING, ADMIRING)
13
EMOTION, DECISION MAKING, AND RISK
Emotions can both enhance and undermine effective decision making. Decision literature (e.g., the Risk as Feeling hypothesis) typically defines emotions simply in term of positive and negative valence. Affective neuroscience has found evidence for specific neurochemical systems underlying a variety of emotions that may be involved in risk/decisions. POSITIVE INDIVIDUALIST (CONFIDENT, SECURE, SATISFIED) NEGATIVE INDIVIDUALIST (NERVOUS, UNSURE, FEARFUL) POSITIVE PROSOCIAL (PROUD, TRUSTING, ADMIRING) NEGATIVE PROSOCIAL (GUILTY, ASHAMED, EMBARRASSED)
14
EMOTION, DECISION MAKING, AND RISK
Emotions can both enhance and undermine effective decision making. Decision literature (e.g., the Risk as Feeling hypothesis) typically defines emotions simply in term of positive and negative valence. Affective neuroscience has found evidence for specific neurochemical systems underlying a variety of emotions that may be involved in risk/decisions. POSITIVE INDIVIDUALIST (CONFIDENT, SECURE, SATISFIED) NEGATIVE INDIVIDUALIST (NERVOUS, UNSURE, FEARFUL) POSITIVE PROSOCIAL (PROUD, TRUSTING, ADMIRING) NEGATIVE PROSOCIAL (GUILTY, ASHAMED, EMBARRASSED) POWER (POWERFUL, VIGOROUS, ENERGETIC)
15
EMOTION, DECISION MAKING, AND RISK
Emotions can both enhance and undermine effective decision making. Decision literature (e.g., the Risk as Feeling hypothesis) typically defines emotions simply in term of positive and negative valence. Affective neuroscience has found evidence for specific neurochemical systems underlying a variety of emotions that may be involved in risk/decisions. POSITIVE INDIVIDUALIST (CONFIDENT, SECURE, SATISFIED) NEGATIVE INDIVIDUALIST (NERVOUS, UNSURE, FEARFUL) POSITIVE PROSOCIAL (PROUD, TRUSTING, ADMIRING) NEGATIVE PROSOCIAL (GUILTY, ASHAMED, EMBARRASSED) POWER (POWERFUL, VIGOROUS, ENERGETIC) Strategy: include 3 or 4 item measures of every emotion that may be associated with responding to a given risk/decision, and analyze to find common factors.
16
EMOTION, DECISION MAKING, AND RISK
As safety-critical applications (e.g., drone operation) often involve risky situations and are likely to evoke complex strong emotions (e.g., fear, uncertainty, trust, suspicion, excitement), if not addressed properly, they may interfere with mindful cognitive processing of risks! Buck, R., & Ferrer, R., Emotion, warnings, and the ethics of risk communication. In S. Roeser, R. Hillerbrand, P. Sandin, & M. Peterson (Eds.), Handbook of Risk Theory (pp ). New York: Springer, 2012. Buck, R., & Davis, W. A., Marketing risk: Emotional appeals can promote the mindless acceptance of risk. In S. Roeser (Ed.), Emotions and Risky Technologies (Vol. 5, pp ). New York: Springer,
17
The GREAT Emotions and Trust
18
The GREAT Emotions: Human-Human Interaction
The GREAT emotions: The Passions of Civility GREAT emotions (Gratitude, Respect, Elevation, Appreciation, and Trust) are continually, albeit unconsciously, displayed in interaction. They only function if expressed freely by all concerned. As long as they are working, they are not noticed, but if the rules of politeness are not followed the emotional result is instant and negative. Buck, R., Emotion: A biosocial synthesis. Cambridge, UK: Cambridge University Press, 2014.
19
The GREAT Emotions: Human-Human Interaction
The GREAT emotions: The Passions of Civility How to emulate GREAT emotions? GREAT emotions (Gratitude, Respect, Elevation, Appreciation, and Trust) are continually, albeit unconsciously, displayed in interaction. They only function if expressed freely by all concerned. As long as they are working, they are not noticed, but if the rules of politeness are not followed the emotional result is instant and negative. Buck, R., Emotion: A biosocial synthesis. Cambridge, UK: Cambridge University Press, 2014.
20
Despite overwhelming evidence supporting the importance of emotion in decision making and development of trust – Current designs of human-computer interactions ignore critical emotional bases of human interaction (e.g., Gratitude, Respect, Elevation, Appreciation, and Trust) while designing human-computer interaction.
21
How should we design effective trust-inducing human-computer interaction?
22
Main Idea In human-human interaction, emotional expressiveness and nonverbal communication are shown to play a major role in long term development and maintenance of trust Persons who are generally more expressive of their feelings and desires are both easier to “read” and tend to engender expressiveness in interaction partners, generating an interaction in which trust can emerge efficiently. Bonnie M. Muir Trust between humans and machines, and the design of decision aids. Int. J. Man-Mach. Stud. 27, 5-6 (November 1987), Riggio, R. E., & Friedman, H. S., Impression formation: The role of expressive behavior. Journal of Personality and Social Psychology, 50, 421 – 427, 1986. Sabatelli, R. M. & Rubin, M., Nonverbal expressiveness and physical attractiveness as mediators of interpersonal perceptions. Journal of Nonverbal Behavior, 10(2), 120 – 133,
23
In principle, expressive communication between human and computing systems can be an effective way to promote trust and address negative emotions (e.g., anxiety, fear, suspicion, doubt)!
24
Key Research Questions
What to communicate? How to communicate to promote trust? How to adapt communication to maintain trust over time?
25
Research Challenge - I What information regarding system performance is relevant to “trust” and need to be communicated in safety-critical scenarios?
26
Information Communication Framework Promoting Trust
Research Challenge - I What information regarding system performance is relevant to “trust” and need to be communicated in safety-critical scenarios? Information Communication Framework Promoting Trust
27
Study Design Participants will be given one of six scenarios in which they are to imagine themselves in the role of a drone operator or system administrator in a high, medium, or low risk drone operation. In the high risk scenario, the drone will be over a battlefield and the decisions involve identifying enemy targets who may be innocent civilians. In the moderate risk scenario, the drone will be over a border region and the decisions will involve arresting or not arresting suspected illegal immigrants who may be innocent citizens. In the low risk operation, the drone will be over the ocean and the decisions involve identifying for whale watchers marine mammals which may be interesting whale pods or non-interesting seals.
28
Emotion and Reason Items
While considering the imagined scenario, participants will rate various aspects of the system performance in terms of their relevance to various thoughts and feelings of the user. Does slow response time make you feel X? Does slow response time make you think about X? Does the occasional communication error make you feel X? Does the occasional communication error make you think about X? Does success make you feel X? Does failure make you think about X? X is “of differences between,” “of pros and cons of,” “of arguments for,” “of facts about,” and “about consequences of.” correctly identifying or not identifying/arresting or not arresting/identifying or not identifying the target?
29
GREAT Emotions Items Do you trust the system to function correctly?
Do you have respect for the system designers? Are you grateful that you have the system? Do you appreciate the system? Are you suspicious of the system?
30
Status of the study The study is approved by the UConn IRB and is currently being reviewed by the AFOSR IRB.
31
Research Challenge - II
How to effectively communicate information regarding system states to end users to promote trust in safety-critical scenarios?
32
After all, it failed in many other domains!
While the need to design some sort of information/warning delivery mechanism seems obvious, it is likely to fail unless designed carefully! After all, it failed in many other domains! For example, in hospitals, management of warning/alarm signals is already a significant problem! Meredith, Christina, and Judy Edworthy. "Are there too many alarms in the intensive care unit? An overview of the problems." Journal of advanced nursing 21.1 (1995): Ignoring security warnings or software update recommendation is pretty common as well! Michael Fagan and Mohammad Maifi Hasan Khan. “Why Do They Do What They Do? A Study of What Motivates Users to (Not) Follow Computer Security Advice”. In proceedings of Symposium On Usable Privacy and Security (SOUPS), 2015.
33
We ourselves are currently investigating the problem of designing effective software update/warning design/delivery mechanisms, which is relevant to this project and presented next! (Please note that the study presented next is funded by NSF award CNS )
34
We are investigating the emotional responses to pop-up warning messages while using computing system and its influence on users’ behavior. The initial risk chosen was responding to pop-up warnings to update software. This is an outwardly unremarkable risk situation: ordinary, routine, even mundane.
35
Study Design Assessed emotions reported to pop-up warnings while pressured or relaxed on the Web (Warning Pop-Up Emotions Scale: WPE). We asked respondents to rate 45 emotions, and used Factor Analysis techniques to discover underlying factors.
36
METHOD 400 participants recruited through Mechanical Turk (209 F, 190 M) in an anonymous online ½ hour Qualtrics survey.
37
METHOD 400 participants recruited through Mechanical Turk (209 F, 190 M) in an anonymous online ½ hour Qualtrics survey. Included a range of ages; educational levels; reported computer proficiency.
38
METHOD 400 participants recruited through Mechanical Turk (209 F, 190 M) in an anonymous online ½ hour Qualtrics survey. Included a range of ages; educational levels; reported computer proficiency. Participants were asked about emotions experienced when pop-up warnings appear while surfing the web with no specific purpose, versus hard at work on an important project with a looming deadline.
39
METHOD 400 participants recruited through Mechanical Turk (209 F, 190 M) in an anonymous online ½ hour Qualtrics survey. Included a range of ages; educational levels; reported computer proficiency. Participants were asked about emotions experienced when pop-up warnings appear while surfing the web with no specific purpose, versus hard at work on an important project with a looming deadline. Two Effort conditions (Relaxed-Pressured or Pressured-Relaxed), and 45 emotions in each, were presented in random order.
40
Key Findings
41
RESULTS- STRUCTURE ANALYSIS
The structure of reported emotions was examined by exploratory structural equation modeling (ESEM).
42
RESULTS- STRUCTURE ANALYSIS
The structure of reported emotions was examined by exploratory structural equation modeling (ESEM). ESEM combines exploratory (EFA) and confirmatory factor analysis (CFA).
43
RESULTS- STRUCTURE ANALYSIS
The structure of reported emotions was examined by exploratory structural equation modeling (ESEM). ESEM combines exploratory (EFA) and confirmatory factor analysis (CFA). Is indicated when prior measurement knowledge is limited.
44
RESULTS- STRUCTURE ANALYSIS
The structure of reported emotions was examined by exploratory structural equation modeling (ESEM). ESEM combines exploratory (EFA) and confirmatory factor analysis (CFA). Is indicated when prior measurement knowledge is limited. In both Effort conditions, the WPE scale factored into four components with 5 indicators each.
45
RESULTS- STRUCTURE ANALYSIS
The structure of reported emotions was examined by exploratory structural equation modeling (ESEM). ESEM combines exploratory (EFA) and confirmatory factor analysis (CFA). Is indicated when prior measurement knowledge is limited. In both Effort conditions, the WPE scale factored into four components with 5 indicators each. Positive: Happy, Confident, Secure, Respectful, Grateful. Reliability: Relaxed = .882; Pressured = .896
46
RESULTS- STRUCTURE ANALYSIS
The structure of reported emotions was examined by exploratory structural equation modeling (ESEM). ESEM combines exploratory (EFA) and confirmatory factor analysis (CFA). Is indicated when prior measurement knowledge is limited. In both Effort conditions, the WPE scale factored into four components with 5 indicators each. Positive: Happy, Confident, Secure, Respectful, Grateful. Reliability: Relaxed = .882; Pressured = .896 Anxious: Confused, Anxious, Nervous, Afraid, Freaked out. Reliability: Relaxed = .880; Pressured = .879
47
RESULTS- STRUCTURE ANALYSIS
The structure of reported emotions was examined by exploratory structural equation modeling (ESEM). ESEM combines exploratory (EFA) and confirmatory factor analysis (CFA). Is indicated when prior measurement knowledge is limited. In both Effort conditions, the WPE scale factored into four components with 5 indicators each. Positive: Happy, Confident, Secure, Respectful, Grateful. Reliability: Relaxed = .882; Pressured = .896 Anxious: Confused, Anxious, Nervous, Afraid, Freaked out. Reliability: Relaxed = .880; Pressured = .879 Lonely: Lonely, Abandoned, Ashamed, Isolated, Humiliated Reliability: Relaxed = .910; Pressured = .923
48
RESULTS- STRUCTURE ANALYSIS
The structure of reported emotions was examined by exploratory structural equation modeling (ESEM). ESEM combines exploratory (EFA) and confirmatory factor analysis (CFA). Is indicated when prior measurement knowledge is limited. In both Effort conditions, the WPE scale factored into four components with 5 indicators each. Positive: Happy, Confident, Secure, Respectful, Grateful. Reliability: Relaxed = .882; Pressured = .896 Anxious: Confused, Anxious, Nervous, Afraid, Freaked out. Reliability: Relaxed = .880; Pressured = .879 Lonely: Lonely, Abandoned, Ashamed, Isolated, Humiliated Reliability: Relaxed = .910; Pressured = .923 Hostile/Contempt: Disdainful, Scornful, Contemptuous, Hostile, Resentful. Reliability: Relaxed = .874; Pressured = .843
50
RESULTS The structure of reported emotions in response to warnings was invariant across situation (Pressured vs. Relaxed) and sex.
51
RESULTS The structure of reported emotions in response to warnings was invariant across situation (Pressured vs. Relaxed) and sex. The levels of reported emotions varied with situation and sex.
52
Women reported higher Anxiety than men (P).
53
Women reported higher Anxiety than men (P).
Men reported higher Hostile Contempt than women (R and P).
54
Women reported higher Anxiety than men (P).
Men reported higher Hostile Contempt than women (R and P). Men reported higher Loneliness than women (R).
55
Women reported higher Anxiety than men (P).
Men reported higher Hostile Contempt than women (R and P). Men reported higher Loneliness than women (R). Men reported higher overall emotion than women (R).
56
Pressured condition produced higher reports of Anxiety, Contempt, Loneliness, and Total emotion.
57
Pressured condition produced higher reports of Anxiety, Contempt, Loneliness, and Total emotion.
Relaxed condition produced higher reports of positive emotion
58
DISCUSSION This was a specific, narrow, and outwardly unremarkable risk situation; and might be considered to be a weak manipulation.
59
DISCUSSION This was a specific, narrow, and outwardly unremarkable risk situation; and might be considered to be a weak manipulation. Nevertheless, the relaxed versus pressured conditions were associated with significant differences in reported emotion.
60
DISCUSSION This was a specific, narrow, and outwardly unremarkable risk situation; and might be considered to be a weak manipulation. Nevertheless, the relaxed versus pressured conditions were associated with significant differences in reported emotion. The reported emotions were highly reliable and demonstrated a robust structure that was invariant across Pressured vs. Relaxed situation and participant sex.
61
DISCUSSION This was a specific, narrow, and outwardly unremarkable risk situation; and might be considered to be a weak manipulation. Nevertheless, the relaxed versus pressured conditions were associated with significant differences in reported emotion. The reported emotions were highly reliable and demonstrated a robust structure that was invariant across Pressured vs. Relaxed situation and participant sex. Significant differences were found due to situation and participant sex, showing evidence of predictive validity. This study underscores the need for development of emotion-aware communication strategies!
62
Plan for Year 2 How to communicate?
63
Year 2 Goals Develop an Interactive System to Study Different Modes of Communication. Interaction Adaptation Is Module Responsible for: Development of Trust Maintenance of Trust User Interaction Adaptation Module Facial Expression Application Usage Behavior (e.g., time spent to respond) User State Inference Module Physiological Sensor Data Data Collection Module
64
Year 2 Goals Experiment with different modes of communication to communicate system health status to users. We will design hypothetical scenarios to emulate various safety-critical situations, and see how users react to various information and mode of communication.
65
Plan for Year 3 Investigating the Effects of Trust Violations and Ways to Recover Trust
66
A recent work by Kreuger et al illustrates the processes underlying the development and maintenance of a trusting relationship using a game scenario in which two persons interacted in a game situation while both were in a brain scanning device. This study employed hyperfunctional magnetic resonance imaging (hyperfMRI) to measure the brain responses of two strangers as they interacted in a sequential reciprocal trust game. Krueger, F., McCabe, K., Moll, J., Kriegeskorte, N., Zahn, R. Strenziok, M., ... & Grafman, J. , Neural correlates of trust. Proceedings of the National Academy of Sciences of the United States of America, 104, 20084–20089, 2007.
67
Pairs were divided into non-defectors, in which neither participant defected, and defectors, in which some defection was experienced by the pair. Trust was higher in the non-defector group and that it increased over stages, while in the defector group trust decreased across stages. Brain activity patterns reflected the role of empathy and the emergence of mutual goodwill and social bonding. Defector pairs appeared to adopt a strategy whereby they actively evaluated the risks and benefits of cooperation. When trust was achieved, there was no longer any need to empathize actively with the partner: they essentially could relax and enjoy the interaction.
68
In the context of safety-critical system operation, we expect the followings –
Initially, system users will be cautious in trusting their systems completely. As trust is achieved, the operator will cease to need to actively think about the operations of the system, freeing the operator to concentrate on the task at hand. Once trust is violated due to system failures, we expect user’s system usage behavior to change.
69
Year 3 Goals: Investigating the Effects of Trust Violations and Ways to Recover Trust
A series of studies will be conducted to investigate the followings: The effect of violations of trust on the development of a trusting relationship between participant and system The effect of different remediation efforts
70
Thank You!
Similar presentations
© 2025 SlidePlayer.com Inc.
All rights reserved.