Recent rapid progress in technology has raised questions about the potential of artificial intelligence (AI) to replace people in certain jobs, many of which would seem to require a moral compass. Do Americans perceive AI as capable of exhibiting moral judgment?
By 54% to 22%, Americans say they don’t think AI could learn to be moral, according to polling late last year as part of a YouGov project on morality. Republicans (65%) are more likely than Democrats (52%) and Independents (49%) to say machines couldn't learn to be moral. Adults 45 and older (60%) are more likely to say than younger adults (46%) to say so.
If they set aside morality, do Americans think that AI robots could be better than humans at certain kinds of decision-making?
When it comes to tech in the justice system, most Americans (56%) say that a person who can use their emotion and instinct would be a better judge in a court of law, rather than an AI robot that works purely on processing data; 19% say the robot would do better. Only 9% of Americans say an AI robot and a person would be equally good judges. A larger share of Republicans (63%) than Democrats (55%) or Independents (52%) say that a person who can use their emotion and instinct would be a better judge. Americans 45 and older (63%) also are more likely than younger adults to say that a person would be a better judge than an AI robot in a court of law.
Like with judges, Americans prefer a person to an AI robot for making hiring decisions: By 56% to 24%, they say that a person who can use their emotion and instinct would be better at scanning resumes for a job than an AI robot that works purely on processing data. Republicans (65%) are more likely than Democrats (55%) and Independents (49%) to choose a person over a roboto for scanning resumes. Americans 45 and older (31%) are more likely than younger adults to prefer a person for this job (65% vs. 45%).
When it comes to the moral issue of automating jobs in factories, American opinion is divided. Almost half (47%) say that automating jobs in factories is sometimes morally acceptable, while only 17% say it is always morally acceptable, and 14% say it is never morally acceptable. While opinions on this do not vary much by political identification, age, or education, higher family income is linked to support for automation: Americans from families with an annual income of $100,000 or more (24%) are more likely than those with an annual family income between $50,000 and $100,000 (18%) and those with an annual income under $50,000 (14%) to say that automating jobs in factories is always morally acceptable.
Automated weapon systems in the military are opposed by most Americans: 58% say it would be morally unacceptable for a military AI robot to have the capability to decide whether to kill someone or not, while only 21% say they find it morally acceptable. Men (27%) and adults under 45 (32%) are more likely to find AI military robots making killing decisions to be morally acceptable than women (16%) and Americans 45 and older (13%). Current and former military personnel are particularly likely (37%) to find it acceptable for AI military robots to decide whether to kill someone, but that percentage is just 15% among Americans who currently have or previously have had family members serving in the military.
— Carl Bialik and Milan Dinic contributed to this article.
Morality and religion: How closely do Americans think they are related?
Do Americans think being moral is the same thing as following the law?
How do Americans view the morality of personal and societal health decisions?
AI doomsday worries many Americans. So does apocalypse from climate change, nukes, war, and more
Which occupations do Americans think will be most affected by advances in AI?
See the results of this YouGov poll
Methodology: This article includes data from two online polls conducted November 17 - 21 and November 18 - 21, 2022 — each among 1,000 U.S. adult citizens. The 2,000 respondents were selected from YouGov’s opt-in panel using sample matching. A random sample (stratified by gender, age, race, education, geographic region, and voter registration) was selected from the 2019 American Community Survey. The sample was weighted according to gender, age, race, education, 2020 election turnout and presidential vote, baseline party identification, and current voter registration status. Demographic weighting targets come from the 2019 American Community Survey. Baseline party identification is the respondent’s most recent answer given prior to March 15, 2022, and is weighted to the estimated distribution at that time (33% Democratic, 28% Republican). The margin of error for the overall sample is approximately 2%.
Image: Pexels (cottonbro studio)