01.Artificial Intelligence in Public Health
10.1.2 Why do we use AI in Public Health?
11.1.3 The Organization of the Book Content
12.1.4 Discussion Questions
16.2.3 Rule-Based Systems (RBS) and Expert Systems (ES)
17.2.4 Data-Driven AI Models
18.2.4.1 Machine Learning (ML)
19.2.4.2 Neural Networks (NNs)
20.2.4.3 Deep Learning (DL)
21.2.4.4 Transformers (T)
22.2.4.5 Mixture of Experts (MoE)
23.2.4.6 FP8 (8-Bit Floating Point)
24.2.4.7 Reinforcement Learning (RL)
25.2.4.8 Large Language Models (LLMs)
26.2.4.9 Natural Language Processing (NLP)
27.2.4.10 Computer Vision (CV)
28.2.4.11 Combined “Rule-Based” and “Data-Driven” Models
29.2.5 Robotics, AI Agent and Networks of AI Agents
31.2.7 Internet and Internet of Things (IoT)
32.2.8 General Development Processes of AI Models
33.2.9 AI Ethical Guidelines and Governance Frameworks
34.2.10 Discussion Questions
35.Part II: Modeling of AI Applications for Public Health
36.3. Design Models for AI in Public Health
38.3.2 Software Development Process Model for Designing AI Applications
39.3.2.1 Waterfall Process Model
40.3.2.2 Project Process Management Methodology
41.3.3 Understand Current Public Health Reality with AI
42.3.3.1 What is Health? Health Outcomes
43.3.3.2 Systemic Factors and Personal Responsibilities
45.3.3.4 Unrealized Needs
46.3.4 Empower Current Public Health Approaches with AI
47.3.4.1 Primary Prevention
48.3.4.2 Secondary Prevention
49.3.5 Develop New AI-Driven Public Health Approaches
50.3.6 Ethical Considerations and Regulations
51.3.6.1 Big Picture of Ethical Issues for AI in Public Health
52.3.6.2 Ethical Issues in the AI Application Level
53.3.6.3 Regulations for AI in Public Health
54.3.6.3.1 Legal Standards and Regulations
55.3.6.3.2 Ethical Standards and Guidelines
56.3.7 Concept-Based AI Application Design Model1
57.3.7.1 Simple Human Cognitive Functions with AI
58.3.7.2 Complex Cognitive Functions with AI
59.3.8 Technology-Based Design Model
60.3.8.1 “Superhuman” Strengths
61.3.8.2 “Non-human” Strengths
62.3.9 Discussion Questions
63.4. Evaluation Methods for AI in Public Health
64.4.1 Introduction to Evaluation Methods
65.4.1.1 Why is Evaluation Important?
66.4.1.2 General Evaluation Approaches
67.4.1.2.1 Objectivist Evaluation Approaches
68.4.1.2.2 Subjectivist Evaluation Approaches
69.4.1.3 General Evaluation Concepts
70.4.2 What to Evaluate for AI Applications for Public Health?
71.4.2.1 Before AI Projects Get Started
72.4.2.2 During AI Design and Development
73.4.2.3 After Deployment and Using AI Applications
74.4.2.3.1 Evaluating AI system functions
75.4.2.3.2 Evaluating usages of AI systems
76.4.2.3.3 Evaluating the Impact of AI Applications for Public Health
77.4.3 Case Studies of Evaluation Designs for AI in Public Health
78.4.3.1 Case Scenario A: Accuracy of AI Chatbots
79.4.3.2 Case Scenario B: Accuracy of AI Doctors
80.4.3.3 Case Scenario C: Accountability of Self-driving Cars
81.4.3.4 Case Scenario D: User experience of AI Mental Health Support
82.4.4 Discussion Questions
83.Part III: Big Data in Public Health
85.5.1 Data Concepts for AI
86.5.1.1 Data Concepts Map
88.5.1.3 Data and AI Development Stages
89.5.2 Data Issues for AI in Public Health
90.5.2.1 Data Availability
94.5.2.5 Data Privacy and Security
95.5.3 Discussion Questions
96.6. Public Health Domain Data for AI
97.6.1 Introduction for Public Health AI Data
98.6.1.1 AI Pipeline Perspective
99.6.1.2 EMR and Meaningful Uses (MU)
100.6.2 Domain Data for Systemic-Factor Public Health AI1
101.6.2.1 Domain Data for Disease Monitoring
102.6.2.2 Domain Data for Epidemiological Forecasting
103.6.2.3 Emergency and Disaster Response
104.6.2.4 Population Health Analytics
105.6.2.5 Social Determinants of Health (SDOH)
106.6.2.6 Resource Management
107.6.2.7 Policy Optimization
108.6.3 Domain Data for Personal-Responsibility Public Health AI2
109.6.3.1 Concept Maps for Personal Health Data
110.6.3.2 Metadata for Personal Health Records
111.6.4 Discussion Questions
112.Part IV: Cases Studies of AI Applications for Public Health
113.7. Al Applications for Public Health Systemic Factors
114.7.1 Design Process of AI Applications in Public Health
115.7.2 Surveillance and Forecasting
116.7.2.1 Disease Monitoring
117.7.2.2 Epidemiological Forecasting
118.7.3 Emergency and Disaster Response
119.7.4 Population Health Management
120.7.5 Health Equity-Aware Risk Stratifications
121.7.6 Discussion Questions
122.8. Al Applications for Public Health Personal Responsibilities
123.8.1 New Concepts for Public Health Personal Responsibility
124.8.2 PMCO-AI Model with AI Engines
125.8.3 Personalized Opportunity Empowered by AI
126.8.4 Personalized Risk Management and Prediction
127.8.5 Personalized Health Education
128.8.6 Behavioral Nudging and Coaching
129.8.7 Policy-Aware Personalized Opportunity (PAPO) Model
131.8.7.2 PAPO-Heatwave AI
132.8.8 Discussion Questions
133.Part V: Living with AI
134.9. From Acceptance to Thinking Partner
136.9.2 New Constructs for AI Adoption Models
137.9.3 Rethinking Trust in AI
138.9.4 A New Model for Human-AI Thinking Partnership
139.9.5 Discussion Questions
140.10. AI and Workforce for Public Health
141.10.1 AI for Current Workforce’s Continuous Education
142.10.2 AI for Public Health Education
143.10.3 AI for Public Health Informatics Education
144.10.4 Directions of Higher Education Reform
145.10.5 Discussion Questions
147.11. Challenges and Opportunities in the Future
148.11.1 Challenges for AI in Public Health
149.11.2 Opportunities for AI in Public Health
151.11.4 Rethinking Higher Education in the Age of AI
152.11.5 Predictions About AI Cultures in Public Health
153.11.6 Discussion Questions
154.Appendix A: A Meta-AI Homework Sample
155.Appendix B: A Meta-AI Homework Development Framework
156.Appendix C: A Grading Rubric for Meta-AI Homework1
157.Appendix D: A List of Case Scenarios for the Evaluation Study Design
158.Appendix E. A List of Case Scenarios for AI Application Design
168.Fig. 2.1: Multilayers of AI ecosystem
169.Fig. 2.2: Why AI now?
170.Fig. 2.3: Expert Systems
171.Fig. 3.1: A big picture for solving public health problems with AI
172.Fig. 3.2: Concept-based definition of AI
173.Fig. 3.3: Learning, understanding, and adaptation
174.Fig. 4.1: A big picture of some concepts related to AI evaluations
175.Fig. 4.2: The effectiveness/impact of an AI application
176.Fig. 4.3: Quasi-legal study design
177.Fig. 4.4: Responsive/illuminative study design
178.Fig. 5.1: Concepts map for AI data
179.Fig. 5.2: The AI development stages
180.Fig. 6.1: Chain reaction for meaningful uses of EMRs
181.Fig. 6.2: A concept map for domain data type for PHR
182.Fig. 6.3: Metadata for AI-powered personal health records
183.Fig. 7.1: A simplified design process of AI applications in public health
184.Fig. 7.2: Equity-aware risk stratification model (EquiRisk)
185.Fig. 8.1: Ranking of PMCO-AI components based on realistic expectations
186.Fig. 8.2: PMCO-AI-Engine: Behaviour change pyramid with AI engines
187.Fig. 8.3: Personalized opportunities in public health
188.Fig. 8.4: The design of the policy-aware personalized opportunity model
189.Fig. 11.1: Urgent challenges of AI in public health
190.Fig. 11.2: Long-time challenges of AI in public health
191.Table 3.1: Some AI project management methodologies
192.Table 3.2: Systemic factors and personal responsibilities
193.Table 3.3: Examples of realized needs for AI applications in public health
194.Table 3.4: Some approaches identifying unrealized needs in public health
195.Table 3.5: The top ten occupations with severe occupational hazards
196.Table 3.6: AI applications in public health based on perception
197.Table 3.7: AI applications in public health based on attention
198.Table 3.8: AI applications in public health based on memory for recall
199.Table 3.9: AI applications in public health based on NLP
200.Table 3.10: AI applications in public health based on reasoning
201.Table 3.11: AI applications based on planning and decision-making
202.Table 3.12: AI mimic functions for emotional processing
203.Table 3.13: “Parallel tasks” in public health for MPP applications
204.Table 3.14: Public health tasks unsuitable for MPP
205.Table 3.15: Successful database projects in medicine and public health
206.Table 3.16: Selective forgetting examples in public health
207.Table 3.17: AI continuous operation without fatigue for public health
208.Table 3.18: AI’s exploration in non-intuitive spaces in public health
209.Table 3.19: Examples of AI multimodal fusion in public health
210.Table 3.20: Positive AI self-modifying and evolutionary functions
211.Table 3.21: Some negative public health risks of self-modifying AI
212.Table 4.1: Evaluation questions for some evaluation concepts
213.Table 4.2: Key evaluation factors during the development
214.Table 4.3: General evaluation factors for AI functional performance
215.Table 4.4: Examples for evaluation factors for NLP functional performance
216.Table 4.5: Evaluation factors of usages for various AI applications
217.Table 4.6: Some evaluation metrics for measuring disease prevention
218.Table 4.7: Some evaluation metrics for “live longer” and “live better”
219.Table 4.8: Part I: Some evaluation metrics for the impacts on organizations
220.Table 4.9: Part II: Some evaluation metrics for the impacts on individuals
221.Table 5.1: The general data content types
222.Table 5.2: The public health data content types
223.Table 5.3: The data case scenario and choices of learning process
224.Table 5.4: Examples of input-output paired data for various AI tasks
225.Table 5.5: Data requirements for pretraining and continual training
226.Table 5.6: Data availability rank for infectious diseases
227.Table 5.7: Data completeness issue for infectious diseases
228.Table 5.8: Key CVD databases and registries with high data validity
229.Table 5.9: Types of data noise
230.Table 5.10: Examples of data noise in social media for CVDs
231.Table 5.11: Examples of relevant data for dementia
232.Table 5.12: Data context dimensions for AI
233.Table 5.13: Examples of context data for dementia
234.Table 5.14: Model complexity for some AI tasks
235.Table 5.15: Data sufficiency issues for AI models
236.Table 5.16: The typical examples of poor data representativeness.
237.Table 5.17: The challenges for SDOH data representativeness
238.Table 5.18: The needs for data privacy and security for substance abuse
239.Table 6.1: Core objectives of MU for public health
240.Table 6.2: Data types related to disease monitoring
241.Table 6.3: Domain data for disease monitoring from AI perspective
242.Table 6.4: Assumptions of the SEIR model
243.Table 6.5: Assumptions in epidemiological forecasting
244.Table 6.6: Data types related to epidemiological forecasting
245.Table 6.7: Data for epidemiological forecasting from AI perspectives
246.Table 6.8: Data for flooding response from AI pipeline perspectives
247.Table 6.9: Data for substance use from AI pipeline perspectives
248.Table 6.10: Data for education from AI pipeline perspectives
249.Table 6.11: Data for vaccine equity from AI pipeline perspectives
250.Table 6.12: Examples of public health policy modeling
251.Table 6.13: Example data for “agent-based” policy modeling
252.Table 6.14: Example data for “reinforcement learning” policy modeling
253.Table 6.15: Metadata for AI-powered personal health records
254.Table 7.1: Assumptions of choices of diseases to monitor
255.Table 7.2: AI strengths and weaknesses for disaster response
256.Table 7.3: Population risk stratification for community vulnerability
257.Table 7.4: Population risk stratification for health event prediction
258.Table 7.5: Functions of equity-aware risk stratification model (EquiRisk)
259.Table 8.1: Examples of PMCO-AI
260.Table 8.2: Real-time pattern recognition in personal AI applications
261.Table 8.3: Explorations of clarity, insight, and companion engines
262.Table 8.4: Key functions of AI empowerment tools
263.Table 8.5: Key functions of AI navigation tools
264.Table 8.6: Real-world cases for the need of “sustain with others.”
265.Table 8.7: Examples of pure policy elements in pre-heatwaves
266.Table 8.8: Examples of static personal data in pre-heatwaves
267.Table 8.9: Examples of policy elements during heatwaves
268.Table 8.10: Examples of real-time personal data during heatwaves
269.Table 8.11: Key AI-powered tasks in personalized opportunity engine
270.Table 8.12: Key AI-powered tasks in opportunity delivery interface
271.Table 8.13: Key AI-powered tasks in feedback stage of PAPO model
272.Table 9.1: Technology acceptance model (TAM)
273.Table 9.2: Unified theory of acceptance and use of technology (UTAUT)
274.Table 9.3: Some constructs for AI-related acceptance models
275.Table 9.4: Transparency level for some common AI models
276.Table 9.5: Some examples of AI explanations
277.Table 9.6: AI thinking partnership model (ATPM)
278.Table 10.1: Data entry and management
279.Table 10.2: Patient appointment scheduling from technology-based analysis
280.Table 10.3: The core values of “AI-informed liberal education”
281.Table 10.4: The key processes of “AI-integrated scientific education”
282.Table A.1: Guided questions for metacognitive learning
283.Table B.1: Meta-AI homework development framework
284.Table C.1: A sample of rubric for grading meta-AI homework for beginners
