AI-Powered Chatbots for Misinformation Management

in Public Emergencies��

​

North Carolina State University

Presented by: Dr. Prof. Yang (Alice) Cheng

June 2025

​

​

Main Challenges in Emergency Communication

• Misinformation and disinformation amplification:�Deepfakes and synthetic text can spread false information quickly, leading to confusion and panic.

Figure 1. Deepfake incidents, 2019-2023. (Kharvi, P. 2024)

Figure 2. AI-generated image of President Donald Trump

(BBC News, 2024)

Main Challenges in Emergency Communication

• Erosion of trust:�As fake content becomes more sophisticated, public trust in legitimate communication channels may decline.

​

Figure 3. Trust in information sources (Edelman Trust Barometer, 2021)

Source: Edelman Trust Barometer, 2025

Main Challenges in Emergency Communication

• Verification bottlenecks:�Emergency responders and media may struggle to verify content rapidly, delaying critical updates.�
• Exploitation by malicious actors: Adversaries can use AI to manipulate narratives or impersonate authorities during crises.�
• Overload of conflicting information: AI-generated noise can drown out accurate and timely messages from credible sources.

​

​

​

​

​

Why Chatbots?

​

​

​

Figure 6. How AI Chatbots work�(Gupta, M. et al., 2023)

• Chatbot: “an instant messaging account that [is] able to provide services using instant messaging frameworks with the aim of providing conversational services to users in an efficient manner” (Rahman et al., 2017, p. 75)

​

• Chatbots can simulate dialogue using natural language and be useful for customer service in different industries (Saglam et al., 2021).

​

​

Situational motivation

User Motivations of Using Chatbots

Information Behavior and Anti-Misinformation Actions

Results

​

​

​

Implications

1. Tailored Messaging�Share clear and accessible information that matches what users are looking for—especially during crises—to reduce confusion and build trust.

​

• Support Proactive Information Sharing�Encourage people to seek and share accurate information, not just block misinformation, to help create a well-informed community.

​

• Design User-Friendly Chatbots�Build chatbots that are easy to use and meet users’ needs—whether they want helpful info, a fun experience, or social interaction.

​

• Use Context to Boost Engagement�Make chatbot responses relevant to the user’s situation or goal to keep them engaged and encourage deeper interaction.

​

How can nonprofit organizations effectively harness chatbot technologies to influence the public’s social media engagement and foster donation intention?

Research Question

Results

Figure 2.

Results of the Structural Model

Practical Implications

Develop chatbots:

• With functional capabilities (e.g., ease of use and usefulness)
• With emotional capabilities (e.g., emotional connection and social presence)
• To improve public trust in their chatbots

​

"Domino effect" scenario:

• Enhanced chatbot features build public trust
• Boost social media engagement
• Lead to greater willingness to donate

​

Importance of crafting chatbot interactions:

• Tailor communication strategies to highlight and enhance functional and emotional features
• Engage content showcasing these aspects is critical for effective online campaigns

​

Public Emergencies

Necessary Ethical and Technical Safeguards

• Human-in-the-loop-protocols:�Require expert oversight for AI-generated content before dissemination.

Figure 5. Human-in-the-loop intervention. (Sanchez, T & Ye, X. 2024)

Thank You!

​

Dr. Cheng�

919-515-1294 | ycheng20@ncsu.edu | North Carolina State University

​

​

Dr. Cheng, as a researcher @Dr. Cheng, as a violinist

​

​

Dr. Cheng, as an instructor Dr. Cheng, as your good friend

​

​

Pin Alice, as a young mother +Alice, if you have a mobile

​

​

​

​