Artificial Intelligence (AI) has been a major event in computer technology for years, primarily because of all its potential applications and perceived pitfalls. However, in the U.S., surveys of the general public's view of AI are watchfully split between optimism and pessimism about AI's role in daily life.
In the book Artificial Intelligence for Dummies (2nd edition), by John Paul Mueller and Luca Massaron, probably one of the most widely consumed books by the U.S. public on that topic, the authors state clearly in their introduction that, "the basis for what you expect from AI is a combination of how you define AI, the technology you have for implementing AI, and the goals you have for AI. Consequently, everyone sees AI differently." Recent surveys seem to bear out the idea that the U.S. public has conflicting feelings about AI—in part because they understand it less well than computer technologists do.
The Ignorant, the Optimists, and the Abstainers
A study released in July by YouGov Plc, a U.K.-based market research and data analytics firm with offices across the world, is the most recent example of how mixed public feeling about AI technology still is. Public Attitudes Toward Artificial Intelligence in the US sought to gain some insight into public perceptions of the AI pillar of the "fourth Industrial Revolution." Beginning in 2023, YouGov began by surveying more than 6 million people about what kinds of questions should be asked in a survey about public attitudes toward AI, then winnowed down those suggestions for a questionnaire the company put to 5,074 U.S. adults, and of which YouGov has now published the results.
The report segments the American public into three groups that YouGov labels the AI "Ignorant" (those who have "limited knowledge" about AI and don't think it very important yet), the AI "Abstainers" (those who have "limited knowledge" about AI but lean toward believing AI "makes fraudulent activity likely"), and the AI "Optimists" (those who "are more likely to know about it and its uses" and "generally believe that AI helps society overall." (Clearly, a general lack of AI knowledge on the part of the respondents seems more than a bit baked in here.)
The report classifies 29 percent of respondents as Ignorant (54 percent of them women, 46 percent men, 58 percent aged 45 or older), 34 percent as Abstainers (60 percent women, 38 percent men, 71 percent 45 or older), and 37 percent as Optimists (39 percent women, 60 percent men, 37 percent 45 or older).
Among those classified as Ignorant, 27 percent expect AI will have a positive impact on the workplace (as opposed to 34 percent of the 5,074 respondents overall), 19 percent positive on law and justice (compared to 27 percent overall), 35 percent positive on education (42 percent overall), and 40 percent positive on healthcare (53 percent overall). Among the Abstainers, only 9 percent expect AI to have a positive influence on the workplace, 7 percent a good effect on law and justice, 16 percent positive on education, and 31 percent positive on healthcare. In marked contrast, 63 percent of Optimists thought that AI would have a positive impact on the workplace, 52 percent positive on law and justice, 72 percent positive on education, and 84 percent positive on healthcare.
Many of the rest of the survey's results are reported only as percentages of all three groups lumped together, and they highlight attitudes that often seem at odds with the results already mentioned. For example, of the 5,074 participants, 61 percent are opposed to AI teaching elementary school lessons, 56 percent opposed to AI teaching high school lessons, and 55 percent opposed to AI teaching university lessons, despite the overall 42 percent approval of AI's potential effect on education in general. What is acceptable to the respondents is AI grading coursework or exams (54 percent), creating lesson plans for humans (56 percent), and compiling academic research (57 percent). The respondents as a whole split at 44 percent each in thinking "reviewing university applications" was both OK and not OK.
The attitude of all respondents about the impact of AI in the field of education gets even murkier when the report breaks down responses to questions about whether a human or an AI would be acceptable in certain educational contexts. When asked if they would "prefer" a human to an AI teacher, 74 percent said "yes" in grade-school settings, 71 percent in high school, and 66 percent in college. Less than 10 percent of all respondents said they'd prefer AI to do this in each setting, and 12 percent (elementary schools), 15 percent (high schools), and 18 percent (college) said they "wouldn't mind" whether teaching was done by a human or AI "as long as it was done well"—without defining anywhere what "well" means.
General Opinions
Moving on to more general results, again lumping all respondents together, the report summarizes a main result as "28 percent of the public trust how AI is currently used," with "39 percent expressing distrust," 26 percent trusting AI's potential future uses and 48 percent distrusting those potential uses. Even among the Optimists, 57 percent express trust in how AI is being used now and barely 51 percent trusting how it will be used in the future. When asked about AI's impact on wider society, the respondents split at 31 percent whether they think AI's benefits outweigh the drawbacks and 31 percent thinking the opposite. Most Optimists (61 percent) think the benefits outweigh the drawbacks, Abstainers (70 percent) think the drawbacks are predominant, and of the Ignorant group, 26 percent say benefits and drawbacks are about equal and 39 percent admit that they just don't know.
Opinions on law and justice had their starkest divides over the issues of using human juries rather than AI for verdicts (75 percent in favor) and use of facial recognition technology (30 percent are OK with either a human or AI doing this, 29 percent think only humans should do this, and 30 percent prefer AI judgment).
In the workplace, 62 percent of all respondents say they favor AI use in the workplace "if a human isn't being replaced," 16 percent said use of AI in the workplace is unacceptable under any circumstances, and eight percent say it's fine to use AI in the workplace even if it results in humans being replaced. Overall, 51 percent said it's acceptable to use AI for training only, 56 percent said it's unacceptable to let AI make decisions about workplace misconduct, and 54 percent object to using AI to screen job applications or requests for promotions and raises (the last two being asked as separate questions).
On the topic of AI's impact on their daily lives, 54 percent of all respondents think AI helps them, just 5 percent think it hinders them, and 37 percent think AI does neither. However, in presenting this data, the report also noted, "in the qualitative groups, there was no consensus as to what constitutes 'AI,' and many of the AI Ignorant and AI Abstainers felt that AI has hindered them. Care and service-oriented processes that were once managed by humans are becoming more automated over time [and] are the most heavily cited source of AI making their lives harder," the report summary notes.
AI developments were not considered a top global issue with all respondents, war and terrorism being cited first (58 percent) while "developments in AI and technology" were tops for only 18 percent, tied with "inequality" in 7th place. Opinion was most divided about AI's impact on the workplace, splitting at 34 percent saying positive, 35 percent saying negative. On the effect on law and justice, 27 percent predict a positive impact and 33 percent a negative impact, with 29 percent thinking it will have neither.
AI's Future in the Public Eye
Sixty-seven percent of all respondents said they think AI will result in fewer jobs overall, 59 percent think it's unlikely their personal job will be replaced by AI, and 29 percent think it likely AI will replace them eventually.
When it comes to creating laws and regulations governing AI, the results offer a couple of small surprises. Although 60 percent overall said there should be further development of AI as long as there were appropriate regulations, 5 percent said there should be no regulation of AI at all, 8 percent said there should be no further development of AI but the regulations currently in place are sufficient, and 9 percent said they thought AI should be completely banned by law and no further development should be permitted. Also of interest were the expressed opinions that some independent body rather than government agencies or legislatures should make regulations, 16 percent saying companies producing AI products should have no input in developing AI laws and regulations, and only 15 percent saying AI product vendors should be involved in that process.
On a happier note, 84 percent of Optimists, 53 percent of respondents overall, and 31 percent of Abstainers all think AI's biggest benefits will come in healthcare. It was the only field in which all three groups agreed on a potential positive impact. Even so, this was tempered by responses to more specific questions. Respondents preferred a human doctor to AI for initial primary-care consultations (72 percent), for prescribing medication (70 percent), making decisions about health-insurance coverage (61 percent), reviewing scans and diagnosing illness (55 percent) and allocating medical resources in hospitals (55 percent). Managing and booking appointments was the only area in which AI held an advantage (39 percent "wouldn't mind" if AI did this, 32 percent would prefer a human, and 23 percent said they'd prefer AI handle it). Even most AI Optimists would prefer a human for initial consultation (60 percent) and prescribing medications (62 percent).
Summaries of Other Surveys About AI
Some other public surveys about AI were carried out in 2023 and earlier. Although some questions such as what public opinion is of the ways and means by which AI functions are glossed over because of the "black box" nature of many AI apps, results are largely similar to those found by the YouGov survey.
A Pew Research Center study published in November 2023 compares data from similar surveys Pew did in 2021 and 2022 and shows the percentages of respondents classifying themselves as "more concerned than excited" about the prospects of AI use in daily life grew from 37 percent in 2021 and 38 percent in 2022 to 52 percent in 2023, while those "more excited than concerned" shrank from 18 percent in 2021 to 15 percent in 2022 and 10 percent in 2023. The 2023 version showed 41 percent of respondents being opposed to using AI for reviewing job applications and 71 percent opposed to using AI to make a final hiring decision. In addition, 61 percent objected to using AI to track worker movements at work, 56 percent objected to using it to track when office workers are at their desks, 51 percent objected to using AI to record exactly what people are doing on their work computers, and 71 percent objected to use of facial recognition technology to analyze employees' facial expressions.
A survey conducted by YouGov for the AI Policy Institute in September 2023 of 1,118 U.S. respondents reported that 63 percent think there should be regulations that prevent AI apps from being able to develop superintelligence. When posed as a question of whether such restrictions might cause the U.S. to fall behind China in AI technology development, 67 percent said there should be restrictions on how powerful AI models can become anyway.
The American Association of Retired Persons (AARP) did a study in November 2023 (updated in May 2024) of respondents aged 50+ about technology in general. To questions about generative AI specifically, 85 percent of respondents are aware of generative AI, 31 percent are excited about potential benefits, and 11 percent think it could enhance their digital safety, but 60 percent are undecided about the technology's impact and just 9 percent say they've tried the technology.
Conclusion: Lots of Inconclusion
Because AI is a relatively new technology that has only achieved widespread public attention since the 2023 unveiling of Chat GPT-4, there aren't that many surveys of public opinion about the technology overall yet. Even this summary shows not all surveys are asking the same questions, all of them may be making different assumptions about what the public considers to be "AI technology," and there's no agreement on important survey aspects such as what constitutes an adequate sample size. While there's nothing like universal condemnation of AI as some kind of threat, there's clearly a lot of work to be done educating the public about what constitutes AI, what its capabilities are, and how confident a still-hesitant public could eventually become in integrating AI into daily life.
LATEST COMMENTS
MC Press Online