Probability Weighting in Interactive Decisions: Evidence for Overuse of Bad Assistance, Underuse of Good Assistance

Title of the Paper

Probability Weighting in Interactive Decisions: Evidence for Overuse of Bad Assistance, Underuse of Good Assistance

Paper Information

• Research Domain: Human-Computer Interaction, Behavioral Economics, Decision Theory in Psychology
• Keywords: Probability weighting, human-computer interaction, assistive user interface, usage bias, prospect theory, design optimization, user decision-making, system reliability

Research Background and Problem

• Identified Problem or Challenge: Assistive user interfaces do not always provide correct assistance. Users often exhibit systematic biases when deciding whether to use assistive functions, which may lead to overuse of low-accuracy systems or underestimation of high-accuracy systems.
• Importance of the Problem: As intelligent systems become increasingly prevalent in interactions, designing systems that enable users to reliably and efficiently utilize assistive functions has become a crucial topic. Misuse or neglect of these functions can significantly reduce user efficiency and satisfaction.
• Research Motivation and Related Work:
  • Studies in economics and psychology have found that individuals exhibit systematic biases when evaluating probabilistic events: low-probability events are overestimated, while high-probability events are underestimated. This phenomenon is modeled using the "Probability Weighting Function."
  • Previous HCI research has revealed user behaviors when interacting with assistive interfaces, such as a preference for low-accuracy interfaces and rejection of high-accuracy interfaces due to loss aversion.
  • Further understanding of how this weighting function manifests in interactive environments is still needed.

Solution

• Proposed Solution: Apply the probability weighting function to quantify biases in user interaction decisions and analyze through experiments whether users' choices regarding assistive functions align with theoretical predictions.
• Innovations:
  1. Established metrics to measure users' overuse and underuse of assistive interface functions.
  2. Clarified the application of the probability weighting function in interactive decision-making and validated its predictive effectiveness.
• Implementation Steps:
  1. Design an experiment manipulating the suitability (accuracy) of assistive interface functions for different task objectives.
  2. Collect user interaction data under varying accuracy conditions to quantify biases.
  3. Fit user behavior data to the probability weighting model.
  4. Propose design recommendations to mitigate decision biases.

Research Findings

• Specific Findings:
  1. Under low accuracy conditions (20%), users tend to overuse assistive functions due to probability weighting that overestimates occasional correct assistance.
  2. Under high accuracy conditions (80%), users tend to underestimate assistive functions and choose not to use them, undervaluing frequent correct assistance.
  3. The probability weighting function fitted to experimental data demonstrates that the model can predict users' decision-making behaviors.
• Advantages Compared to Existing Solutions:
  • Provides a bias explanation and quantification model based on behavioral economics theory, offering stronger theoretical support and predictive power compared to existing HCI methods relying on satisfaction and preference analysis.
  • Highlights the sources of decision biases and their impact on design, offering new perspectives for improving interactive design.
• Experimental or Evaluation Results:
  1. Under low accuracy conditions, users tend to overuse assistive functions due to overestimating low-probability correct assistance (Bias > 0).
  2. Under high accuracy conditions, users tend to underuse assistive functions due to underestimating high-probability correct assistance (Bias < 0).
  3. Experimental results align perfectly with the probability weighting function predictions from prospect theory.
• Limitations and Future Directions:
  1. The experiment only covers static, explicit choice scenarios; further exploration is needed to determine whether similar biases exist in dynamic and implicit decision-making contexts.
  2. The study focuses on specific interaction tasks (text selection); further investigation is required for other forms of interaction.
  3. Suggest developing systems that provide real-time feedback, such as displaying potential benefits of optimized options or explaining the advantages of assistive functions to help reduce decision biases.

Summary and Design Recommendations

• In summary, the study reveals systematic biases in users' utilization of assistive interfaces and explains the origins of these biases using probability weighting theory. Designers can reduce biases by enhancing users' awareness of system performance (e.g., introducing performance monitoring tools). Additionally, to address the issues of overusing incorrect assistance and underusing correct assistance, further optimization of assistive interface design is needed to balance user bias behaviors, thereby improving interaction efficiency and user experience.