Pew Research HOSA BD
Pew Research Center Note
Page 1
60% of Americans uncomfortable with AI in their healthcare
Page 3
Study on Americans' views of AI in health and medicine
Surveyed 11,004 U.S. adults from Dec. 12-18, 2022
Participants from American Trends Panel (ATP)
Survey weighted to be representative of the U.S. adult population
Page 4
Americans uncomfortable with AI in their healthcare
Majority unconvinced AI would improve health outcomes
Survey conducted Dec. 12-18, 2022
Page 5
Americans see both positives and negatives of AI in healthcare
Concerns about impact on patient-provider relationship and security of health records
Majority concerned about the pace of AI adoption in healthcare
Page 6
Caution prevalent in public views on AI in healthcare
Majority concerned about healthcare providers moving too fast with AI adoption
Concerns shared across different groups in the public
Page 7
Majority of U.S. adults uncomfortable with health care provider relying on AI
46% of men comfortable with AI in health care, 54% uncomfortable
66% of women uncomfortable with AI in health care
Higher education, income, and younger adults more open to AI in health care
Discomfort with own health care provider relying on AI prevalent across all groups
Page 8
Those familiar with AI more comfortable with AI in health care
50% comfortable among those who know a lot about AI
63% uncomfortable among those who know a little about AI
70% uncomfortable among those who know nothing about AI
Page 9
Modest share of Americans see AI improving patient outcomes
38% believe AI in health care would lead to better patient outcomes
Men, younger adults, and higher education levels more positive about AI impact
Those familiar with AI more optimistic about AI impact on patient outcomes
Page 10
Americans anticipate positive and negative effects of AI in health care
40% believe AI would reduce mistakes by health care providers
57% expect patient-provider relationships to deteriorate with AI use
Concerns about data security and quality of care with AI in health care
Public divided on AI's impact on quality of care for individuals
Source: PEW RESEARCH CENTER, Survey conducted Dec. 12-18, 2022.
Page 11
Majority of Americans see bias based on race or ethnicity in health and medicine as a major or minor problem
Black adults are especially likely to view bias as a major problem
Smaller share of White adults see bias as a major problem
Survey shows percentages of U.S. adults who perceive bias and unfair treatment based on race or ethnicity in health and medicine
Black adults are most concerned about bias based on patients' race or ethnicity
Page 12
Those who see bias in health and medicine believe AI can improve the situation
Majority think increased use of AI would help reduce bias and unfair treatment
Larger shares among different racial groups believe AI would make bias better than worse
Black adults are slightly less optimistic compared to other groups
Reasons for optimism include AI being more objective and neutral than humans
Page 13
People believe AI can improve bias in health and medicine by being more neutral and consistent
Half of those who see bias believe AI can help address the issue
Survey explores reasons why people think AI can improve bias in health and medicine
Page 14
Some believe bias in health and medicine would stay the same with AI due to existing biases in AI design and data
Others think AI would make bias worse due to reflecting human bias and the need for human judgment in medicine
Page 15
Survey explores views on specific applications of AI in health and medicine
Americans are more open to AI-based skin cancer detection compared to other AI-driven applications
Public awareness of AI in health and medicine is still developing
Note: The survey was conducted from Dec. 12-18, 2022, by the PEW RESEARCH CENTER.
Page 16
AI-based skin cancer screening
AI can scan images of skin to detect potential cancer areas
Majority of U.S. adults (65%) want AI for their skin cancer screening
55% believe AI would make skin cancer diagnoses more accurate
Only 13% think it would lead to less accurate diagnoses
Americans view AI in skin cancer screening as a medical advance
52% see it as a major advance
27% consider it a minor advance
7% say it is not an advance for medical care
Page 17
Demographics and AI in skin cancer screening
Men, younger adults, and those with higher education levels are more enthusiastic about AI screening
Men (72%) more likely than women (58%) to want AI in skin cancer screening
Black adults (57%) less likely than White (65%) and Hispanic (69%) adults to want AI for screening
Younger adults and college graduates more open to AI in screening
Awareness of AI in skin cancer screening increases willingness for AI use
Page 18
AI for pain management recommendations
AI used to help prescribe pain medication and minimize addiction risks
31% of Americans want AI for pain management, while 67% do not
Views on AI-based pain management: 26% say it would improve, 40% see little difference, 32% think it would worsen
Familiarity with AI-based pain management influences openness to AI in care plan
Those aware of AI in pain management more likely to consider AI in their own care
Page 19
Familiarity with AI-based pain management
Those familiar with AI-based pain management more open to using AI in their care plan
47% of those familiar would want AI-based recommendations in post-op pain treatment
Demographic differences in willingness for AI in pain treatment are modest
Majority of most groups do not want AI to help decide their pain treatment
Source: PEW RESEARCH CENTER Survey conducted Dec. 12-18, 2022.
Page 20
AI-driven robots in development for surgical procedures
Expected to increase precision and consistency
Americans cautious about using surgical robots for their own care
40% would want AI-based robotics for their surgery
59% would not want this
Public familiarity with AI-based surgical robots higher than other health applications
59% have heard at least a little about this development
Page 21
Views on AI-driven robots in surgery based on awareness
Unfamiliar individuals more likely to not want them in their care
Men more inclined than women to want AI-based robots for surgery
Higher education levels more open to this technology
Little difference in views between age groups
Preferences for AI applications vary across demographic groups
Page 22
AI chatbots for mental health support using AI
Public reactions to AI chatbots for mental health support are negative
79% of U.S. adults would not want to use an AI chatbot for mental health support
Support for limits on availability or use of AI chatbots for mental health
46% say chatbots should only be used by people seeing a therapist
28% say chatbots should not be available to people at all
23% say chatbots should be available regardless of therapy
Page 23
Majority of U.S. adults across groups lean away from using AI chatbots for mental health support
Even among those familiar with chatbots, 71% would not want to use one for mental health support
Public opinion on AI chatbots for mental health support still developing
Page 24
Acknowledgments to The Pew Charitable Trusts for the report
Primary research team and contributors listed
Project guidance from Pew Research Center methodology team and feedback from other members
Methodology Overview
The American Trends Panel (ATP) by Pew Research Center is a nationally representative panel of randomly selected U.S. adults.
Panelists participate via self-administered web surveys.
Panelists without internet access are provided with a tablet and wireless internet.
Interviews are conducted in English and Spanish.
Managed by Ipsos.
Data in the report is from a panel wave conducted from Dec. 12-18, 2022.
Response rate was 88% with 11,004 out of 12,448 sampled panelists responding.
Margin of sampling error for the full sample is +/- 1.4 percentage points.
Panel Recruitment
ATP created in 2014 with the first cohort invited after a national random-digit-dial survey.
Additional recruitments in 2015 and 2017.
Switched to address-based recruitment in August 2018.
Panelists are removed if inactive or request to be removed.
Address-Based Recruitment
Invitations sent to a stratified, random sample of households from the U.S. Postal Service’s Delivery Sequence File.
Incentives offered for completing surveys.
Subset of adults received follow-up mailings with incentives to join the ATP.
23,176 adults invited, with 20,341 agreeing to join the panel.
Sample Design
Target population: non-institutionalized persons ages 18 and older in the U.S.
All active panel members invited to participate in the survey wave.
Questionnaire Development and Testing
Questionnaire developed by Pew Research Center in consultation with Ipsos.
Rigorously tested on PC and mobile devices before launching the survey.
Incentives
Post-paid incentives offered to all respondents.
Amounts ranged from $5 to $20 based on the group's survey response propensities.
Data Collection Protocol
Field period: Dec. 12-18, 2022.
Postcard experiment conducted among ATP non-tablet household panelists.
Two separate launches: Soft Launch and Full Launch.
Invitations sent via email and SMS with reminders.
Invitation and Reminder Dates
Soft Launch: Dec. 12, 2022
Full Launch: Dec. 13, 2022
Reminders sent on Dec. 15 and Dec. 17, 2022.
Page 28
Data quality checks:
Researchers checked for patterns of satisficing like leaving questions blank or always selecting the first or last answer.
Eight ATP respondents were removed due to this checking.
Weighting process:
ATP data is weighted in a multistep process considering sampling stages and nonresponse.
Base weights are adjusted for changes in recruitment survey design.
Weights are calibrated to align with population benchmarks for nonresponse correction.
Additional weighting dimensions are applied within Black adults.
Page 29
Calibration of weights:
Weights are calibrated to align with population benchmarks and trimmed at 1st and 99th percentiles.
Sampling errors and statistical significance tests consider the effect of weighting.
Sample sizes and margins of error:
Unweighted sample sizes and margins of error at the 95% confidence level for different groups are provided.
Page 30
Dispositions and response rates:
Final dispositions for ATP Wave 119 are listed.
Cumulative response rates and weighted response rates are detailed.
Adjusting income and defining tiers:
Income tiers are created based on adjusted 2021 family incomes.
Upper-, middle-, and lower-income tiers are defined.
Page 31
Adjusting income and defining income tiers:
Upper-, middle-, and lower-income tiers are created based on adjusted family incomes.
Income ranges for each tier are specified.
Note about Asian adult sample:
The survey includes a sample of 371 English-speaking Asian adults.
Views of Asian adults are reported but may not be fully representative.
Page 32
Additional charts and tables:
Majority across groups would not want to use an AI chatbot for mental health support.
Data on U.S. adults' willingness to use AI chatbots for mental health support is presented.
Note: The information provided is based on a survey conducted by the Pew Research Center.
Page 33
Survey conducted by PEW RESEARCH CENTER on American Trends Panel Wave 119
Questions on risk-taking behavior, comfort with AI in healthcare, creativity, and technology use
Participants were asked to rate how well certain phrases describe them
Participants were asked about their comfort level with AI in healthcare
Ratings were provided on comfort with taking risks, being a creative thinker, and enjoying new technology
Page 34
Participants were asked about the potential outcomes of using AI in health and medicine
Views on whether AI would lead to better or worse health outcomes
Concerns about the speed of adoption of AI in healthcare
Survey included questions on bias and unfair treatment in health based on race or ethnicity
Participants were asked about the extent of bias in healthcare based on race or ethnicity
Page 35
Participants were asked about the impact of AI on healthcare quality, mistakes, patient-provider relationships, fairness in treatment, and data security
Views on how AI would affect bias and unfair treatment based on race or ethnicity in healthcare
Ratings were provided on the quality of healthcare, mistakes made by providers, patient-provider relationships, fair treatment for all races and ethnicities, and patient data security
Page 36
Reasons for the belief that bias and unfair treatment based on race or ethnicity would improve with AI use in healthcare
Factors included the neutrality of AI, absence of human bias, and AI's ability to make accurate decisions
Reasons for the belief that bias and unfair treatment based on race or ethnicity would worsen with AI use in healthcare
Factors included AI reflecting human biases, concerns about inequality, and potential reinforcement of biases by medical providers
Page 37
Reasons for the belief that bias and unfair treatment based on race or ethnicity would remain the same with AI use in healthcare
Factors included existing biases in people, training, or data, skepticism about AI's impact on bias, and the belief that racial and ethnic bias is not a significant issue in healthcare.
Page 38
AI for Skin Cancer Detection
AI used to review images of skin with and without cancer to detect patterns
Survey on awareness and opinions about AI for skin cancer detection
Questions on familiarity, perception of advancement, and personal preference for AI in screening
Page 39
AI Chatbots for Mental Health Support
AI chatbots designed to support mental health by responding to users' texts
Survey on awareness and opinions about AI chatbots for mental health support
Questions on familiarity, perception of advancement, personal preference for using AI chatbots, and opinions on their availability
Page 40
AI for Pain Medication Management
AI used to determine pain medication dosage post-surgery to prevent abuse
Survey on awareness and opinions about AI for pain medication management
Questions on familiarity, perception of advancement, personal preference for AI involvement, and impact on treatment
Page 41
AI for Pain Medication Management (cont.)
Survey continues on opinions about AI's role in pain medication management
Questions on perception of AI's impact on treatment effectiveness
Page 42
AI Surgical Robots
Development of AI-powered robots for performing surgical tasks
Survey on awareness and opinions about AI surgical robots
Questions on familiarity, perception of advancement, personal preference for AI surgical robots
Overall, the survey covers public awareness, perceptions, and preferences regarding the use of artificial intelligence in skin cancer detection, mental health support through chatbots, pain medication management, and surgical procedures.
Page 43
AI in Biology
AI used to predict protein structures in cells for developing medical treatments.
Survey on awareness of AI in predicting protein structures.
AI in Agriculture
AI used to recommend plant varieties for drought and heat resistance in crops.
Survey on awareness and perception of AI in producing drought and heat-resistant crops.
Page 44
AI in Weather Forecasting
AI used to predict extreme weather occurrences like heavy rainfall and storms.
Survey on awareness and perception of AI in predicting extreme weather events.
Page 45
AI in News Reporting
AI used to write news articles based on news events and information.
Survey on awareness and perception of AI in news article writing.
AI in Visual Arts
AI used to generate visual images from keywords, including complex artistic images.
Survey on awareness and perception of AI in producing visual images from keywords.