Exciting, Useful, Worrying, Futuristic: Public Perception of Artificial Intelligence in 8 Countries

1 INTRODUCTION

• AI’s growing influence has raised economic, political, social, and ethical questions, including about work, human rights, autonomous weapons, and dual-use technologies.These issues feature in media, academic, industrial, regulatory, and public discussion.Source: §1 · p. 1
• Public-perception research has used surveys, sentiment analysis, and narrative analysis, but much has focused on Western, English-speaking contexts.The paper extends this work with 10,005 respondents across eight countries and six continents.Source: §1 · p. 1
• The survey reports widespread expectations of significant societal impact, support for responsible AI development, and four country-distinguishing sentiment themes.The themes are exciting, useful, worrying, and futuristic.Source: §1 · pp. 1–2

2 BACKGROUND

• AI has no consensus definition, and the paper adopts a similarly broad scope while noting the AI effect and limited public awareness of algorithmic processing.It also shares Nilsson’s definition of AI for comparison with participants’ responses.Source: §2 · p. 2

2.1 Empirical Studies

• Earlier studies, mostly from Western English-speaking settings, report generally positive expectations alongside mixed emotions, social concerns, and support for careful AI management.Other qualitative work finds perceptions of algorithmic systems vary by individual and platform, and users may misunderstand how they operate.Source: §2.1 · p. 2

2.2 Narratives and Media Sentiment Analysis

• Prior analyses describe English-speaking Western AI narratives as tending toward utopian or dystopian extremes, while newspaper coverage has been more optimistic than pessimistic.Studies also report narrative responses that elicited more concern than excitement and findings that did not support a generally negative-news-coverage hypothesis.Source: §2.2 · p. 2

2.3 National Considerations

• National AI strategies and country-specific opportunities and challenges may shape public perception, and researchers have called for greater attention to developing-country perspectives.The cited discussion motivates examining local context in AI development and use.Source: §2.3 · p. 2

3 METHODOLOGY

• The survey was conducted with Ipsos in July 2019 and included 10,005 respondents in eight countries.The work is framed as public-opinion polling.Source: §3 · p. 2

3.1 Instrument Development and Translation

• The instrument combined closed-form and open-ended questions, left AI undefined, and addressed expected social impact, feelings about AI, and country variation.Question development drew on experts, published work, prior research, and a 2018 US pilot; translation used style guides and iterative review.Source: §3.1 · p. 3

3.2 Deployment

• Country panels represented online populations, with stratified recruitment and age and gender quotas; online panels are less representative of general populations where technology access is limited.The survey median length was 23 minutes, and respondents received market-standard incentives.Source: §3.2 · p. 3

3.3 Data Processing and Analysis

• Researchers removed 7.3% of responses for quality and applied modest post-collection weighting adjustments to improve sample representativeness within each country.Removal grounds included bot detection, speeding, and nonsensical or profane open-ended answers.Source: §3.3.1 · p. 3

3.3.1 Quality Checks.

• Professional language-specialized coders and iterative hierarchical checks were used to improve consistency and verify coding across countries.Researchers and the market research partner sampled responses in every country and checked all instances of selected codes.Source: §3.3.2 · p. 3

3.3.2 Coding of Open-Ended Responses.

• For AI-related feelings, 92 codes were grouped into four sentiments or Other, and responses could belong to multiple groups based on assigned codes.Open-ended statistics apply to the specific question, while illustrative quotes may come from any of the four open-ended questions; participants who had heard of AI were designated AI-aware.Source: §3.3.2 · pp. 3–4, §3.3.3 · p. 4

3.3.3 Analysis.

• The analysis anticipated meaningful perception differences associated with development level and included Human Development Index rank in its country comparisons.The paper distinguishes developed-country examples from developing-country examples in describing this analysis.Source: §3.3.3 · p. 4

3.4 Limitations

• Online panels are not representative of the general population, particularly in Brazil, India, and Nigeria, and the research team’s Western residence may limit interpretive context or nuance.The authors also note survey risks such as misunderstanding, translation problems, satisficing, plagiarism, and possible social-desirability bias in Indian panels.Source: §3.4 · p. 4

4.1 Description and Exposure

• Respondents commonly described AI as computers or robots that think independently, learn, or perform human tasks, and reported learning about it chiefly through media and personal networks.The five leading sources were social media (45%), TV reports and commentaries (42%), movies or TV shows (40%), magazine or online articles (32%), and family and friends (31%); personal product experience was also reported (13%). Some respondents showed partial knowledge, associated AI with the Internet or IoT, described it as non-technological, or did not know.Source: §4.1 · p. 5

4.2 Widespread Expectation That AI Will Change The World

• 1-2% (M=1%) believed AI would have little long-term effect on society, indicating that respondents across all eight countries expected significant change.The accompanying table covers long-term societal-impact opinions among respondents who reported being aware of AI.Source: §4.2 · p. 5, Table 2 · p. 6
• Across countries, 56% expected better healthcare versus 12% worse; 52% expected less privacy versus 23% more, while 53% expected more jobs lost versus 22% created.Overall, 46% expected weaker personal relationships versus 22% stronger; in South Korea, 61% expected relationships to weaken.Source: §4.2 · pp. 5–6
• 29% expected AI to be mostly good for society, while 41% said its long-term effect would depend on what happens and 12% expected roughly equal good and bad effects.Mostly-good expectations were 37-51% in Brazil, India, and Nigeria, compared with 18-23% in Australia, Canada, the US, South Korea, and France.Source: §4.2 · p. 6
• 82% considered it very or extremely important for technology companies to weigh pros and cons before release, and 81% to follow a responsible AI-product development process.Respondents also said AI’s effects depend on whether it is used responsibly.Source: §4.2 · p. 7

4.3 Sentiment Groups

• Futuristic was the largest sentiment group (24.4%), followed by Worrying (22.7%), Exciting (18.9%), and Useful (12.2%); 34.9% of responses were coded None.Exciting responses expressed positive enthusiasm, Useful responses described AI as helpful, Worrying covered concerns and fear, and Futuristic included references to advanced technology, science fiction, or the future.Source: §4.3 · p. 7
• Four countries most often found AI Worrying, three found it Exciting, and South Korea most often found it Useful; South Korea had 19% Useful and 38% Futuristic responses.Australia, Canada, and the US were Worrying-led, followed by Futuristic; France had similar Useful, Worrying, and Futuristic shares but fewer Exciting responses. South Korea had the lowest Exciting share (6%) and respondents often mentioned AI assistants and home automation.Source: §4.3 · p. 7
• Exciting dominated in developing countries; India led all countries in Exciting and had second-highest Useful and lowest Worrying, while Brazil showed relatively higher Worrying and Futuristic.Respondents could be coded into multiple sentiment groups.Source: §4.3 · p. 7, Table 3 · p. 8
• 49% of 1802 Exciting respondents expected AI to be mostly good for society, versus 13% of 2129 Worrying respondents.Sentiment groups and expected societal impact overlapped, but neither fully accounted for responses in the other.Source: §4.3 · pp. 7, 9, Figure 2 · p. 9

5 DISCUSSION

• The authors call for investment in responsible design, AI information campaigns, valued offerings, and research on media narratives shaping expectations and sentiment.They connect public expectations that AI will shape the future with a need for design and ethics resources.Source: §5 · p. 9

5.1 AI Design and Ethics Guidance

• AI design guidance and ethical toolkits offer practical strategies for responsible development and deployment, and the authors argue these resources warrant greater priority and use.The findings also reinforce the importance of minimizing potential harms and sharing information about responsible-development efforts.Source: §5.1 · p. 9

5.2 Interventions and Communications

• The authors identify opportunities for education and communications about AI, including emphasizing healthcare applications and maintaining attention to privacy in design and development.They also suggest studying conditions behind South Korea’s unusually strong perception of AI as Useful.Source: §5.2 · p. 9

5.3 Narratives and Sentiment Groups

• Concern about AI is more prominent in the US, South Korea, and other countries named in the discussion than in Brazil, India, and Nigeria.The passage says concern is higher in the US, Australia, and France, and less prominent in South Korea, Brazil, India, and Nigeria.Source: §5.3 · p. 10
• Exciting, Useful, Worrying, and Futuristic sentiment groups vary across countries and offer a more nuanced alternative to classifying responses as simply positive or negative.The authors propose further study of media and pop-culture narratives, country factors, and how sentiment groups relate to technology adoption and views on funding and regulation.Source: §5.3 · p. 10

• The survey included 10,005 respondents across eight countries and six continents and examined expected AI impacts, sentiment, and country variation.Source: §6 · p. 10
• Respondents widely perceive that AI will significantly affect society, but the overall nature of those effects is not yet determined, underscoring responsible development and use.Source: §6 · p. 10
• The paper identifies Exciting, Useful, Worrying, and Futuristic sentiment groups whose prevalence differs across countries, and highlights future research and communication opportunities.Source: §6 · p. 10

• The instrument asks respondents how much they know about AI and invites unaided descriptions and examples of its present-day uses.Source: A QUESTIONNAIRE – SELECT QUESTIONS · p. 11
• Exposure questions ask how much respondents heard about AI in the past year and where they learned about it, with response options spanning media, personal contacts, education, and technology use.Source: A QUESTIONNAIRE – SELECT QUESTIONS · pp. 11–12
• Future-impact questions ask what AI may change in the next decade, using paired options that include jobs, healthcare, relationships, and privacy.Source: A QUESTIONNAIRE – SELECT QUESTIONS · p. 12
• Other items ask about AI’s long-term effect on society, the importance of responsible technology-company practices, feelings about AI, and uncomfortable experiences with AI-related technology.Source: A QUESTIONNAIRE – SELECT QUESTIONS · pp. 11–12