Research Analysis
Tell Me What I Missed: Interacting with GPT during Recalling of One-Time Witnessed Events
Suifang Zhou (City University of Hong Kong), Qi Gong (Shanghai Jiao Tong University), Ximing Shen (Keio University Graduate School of Media Design), and RAY LC (City University of Hong Kong).
LLM-assisted technologies are increasingly used to support cognitive processing and information interpretation, yet their role in aiding memory recall—and how people choose to engage with them—remains underexplored. We studied participants who watched a short robbery video (approximating a one-time eyewitness scenario) and composed recall statements using either a default GPT or a guided GPT prompted with a standardized eyewitness protocol. Results show that default-condition participants who believed they had a clearer understanding of the event were more likely to trust GPT's output, whereas guided-condition participants showed stronger alignment between subjective clarity and actual recall. Additionally, participants evaluated the legitimacy of the individuals in the incident differently across conditions. Interaction analysis further revealed that default-GPT users spontaneously developed diverse strategies, including building on existing recollections, requesting potentially missing details, and treating GPT as a recall coach. This work shows how GPT-user interplay subconsciously affects beliefs and perceptions of remembered events.
Executive Impact: Key Findings at a Glance
Understanding the subtle influences of AI on human memory and perception is crucial for responsible AI design, especially in high-stakes contexts like evidence documentation.
0
Participants Studied
0.0
Prompted GPT Avg. Interactions
0.0
Default GPT Intruder Disfavor
0.00
Clarity-Trust Corr. (Default GPT)
Enterprise Process Flow
Watch 36-second Incident Video
→
15-Minute Memory Formation Break
→
GPT-Assisted Witness Statement Writing
→
60-Minute Memory Consolidation Break
→
Post-Experiment Assessment
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
Memory Recall & Accuracy
Perceptions & Beliefs
Interaction Strategies
Implications & Future Work
Impact on Memory Recall Accuracy
While GPT assistance improved efficiency in statement composition, the study found no statistically significant difference in factual recall accuracy between the default (M=24.23) and guided (M=24.20) GPT conditions. However, the *relationship* between self-perceived clarity and objective recall diverged significantly. In the guided GPT condition, participants' subjective sense of clarity was positively correlated with their actual memory recall performance (r=.55, p=.034), meaning those who felt clearer genuinely performed better. Conversely, in the default GPT condition, higher perceived clarity was correlated with increased trust in GPT's output (r=.66, p=.014), regardless of actual accuracy, suggesting a potential for confirmation bias.
Shaping User Perceptions and Trust
The study revealed that GPT's involvement subtly influenced participants' subjective interpretations of the witnessed event. Notably, default-GPT users reported significantly greater disfavor toward the intruder (M=8.15) compared to guided-GPT users (M=10.00), implying that unstructured interactions might amplify affective biases and lead to more emotionally charged judgments. Participants' prior beliefs about LLM capabilities also dictated their interaction approach, ranging from full reliance on GPT as an authority to using it merely for linguistic refinement, highlighting varied trust levels and potential resistance to AI assistance.
Diverse User Interaction Strategies
In the default GPT condition, participants spontaneously developed a range of strategies for memory recall and statement composition. These included using GPT as a co-writer to structure their recollections, leveraging it to elicit potentially missing details ("tell me what I missed" pattern), and even employing it as a meta-level recall coach to improve their own memory retrieval techniques. One participant noted, "GPT took me along and finished the story while conceptualizing the details, so it's like this kind of very rational and curved-surface way of summarizing the story." In contrast, the guided GPT condition, with its structured eyewitness protocol, naturally led to a more step-by-step elaboration of incident details.
Responsible AI Design & Future Research
This research underscores the inherent fragility of human memory and its susceptibility to LLM influence, especially critical in high-stakes, evidence-based fields like legal documentation and journalism. The findings advocate for cautious AI design that balances efficiency with factual accuracy, mitigating potential biases and user distrust. Future work should explore the absolute effect of GPT by introducing a no-GPT baseline, investigate longer memory consolidation intervals (e.g., 1 week), and employ alternative measurement methods to avoid contaminating memory during assessment.
7.7 Rounds
Average Interactions in Prompted GPT (vs. 2.8 for Default)
| Feature | Default GPT Condition | Guided GPT (Prompted) Condition |
|---|---|---|
| Factual Recall Accuracy | Comparable to Guided GPT (M=24.23) | Comparable to Default GPT (M=24.20) |
| Legitimacy Judgments (Disfavor toward Intruder) |
Significantly higher disfavor (M=8.15, SD=2.70) | Lower disfavor (M=10.00, SD=1.85) |
| Clarity-Output Trust Relationship | Higher self-perceived clarity correlated with trusting GPT output (r=.66, p=.014) | Higher self-perceived clarity correlated with actual recall accuracy (r=.55, p=.034) |
| Interaction Style | Spontaneous strategies: co-writing, eliciting missing details, recall coaching | Structured, step-by-step elaboration based on eyewitness protocol |
Participant Insights: GPT as a "Recall Coach"
In the Default GPT condition, participants often treated GPT as a "recall coach" or a tool to prompt deeper recollection. One participant (P6) explained: "Okay, I watched this happen, like in a very simple description, what are some tips and tricks you can give me on recalling these events or writing a report on what happened? And it gave me a list-focus on setting, what time, etc. And I was like, 'That makes sense. And I just immediately went to writing. So honestly, I think I started with the intent to just use it as a recall coach to begin with.'" This highlights users' proactive efforts to leverage AI not just for direct content generation but also for meta-cognitive support in memory retrieval, demonstrating a flexible and adaptive human-AI collaboration.
Calculate Your Potential AI Impact
Estimate the efficiency gains and cost savings your organization could achieve by implementing responsible AI solutions for documentation and knowledge recall.
Estimated Annual Savings
$0
Annual Hours Reclaimed
0
Your AI Adoption Roadmap
A structured approach to integrating AI for enhanced memory recall and factual documentation, ensuring ethical and effective deployment.
Phase 1: Discovery & Strategy
Assess current documentation workflows, identify key memory recall challenges, and define specific AI objectives. Develop a tailored strategy for ethical AI integration.
Phase 2: Pilot & Prototyping
Implement a small-scale pilot project with a guided GPT assistant, gathering user feedback on interaction patterns and perceived accuracy. Refine protocols based on initial findings.
Phase 3: Rollout & Training
Scale the AI solution across relevant departments, providing comprehensive training on best practices for human-AI collaboration and bias mitigation in memory-intensive tasks.
Phase 4: Monitoring & Optimization
Continuously monitor AI performance, track user perceptions, and refine prompts and guidelines to ensure ongoing accuracy, reliability, and user trust in documentation processes.
Ready to Transform Your Enterprise's Knowledge Management?
Leverage cutting-edge AI insights to improve documentation, enhance recall accuracy, and build trust in your AI-driven workflows. Schedule a personalized consultation to explore how our expertise can benefit your organization.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat