Public trust in AI is low now- A regulatory framework of transparency is key to trust, development and implementation.
Gillespie, Nicole, et. al (2021); (Professor of Management; KPMG Chair in Organizational Trust, The University of Queensland, research fellow; University of Queensland, KPMG; “Executive Summary--Trust in Artificial Intelligence: a five country study”; https://www.researchgate.net/profile/Caitlin-Curtis/publication/356842206_Trust_in_artificial_Intelligence_a_five_country_study/links/61b0342cc2e267424d0fd68f/Trust-in-artificial-Intelligence-a-five-country-study.pdf//EF & acd)
Realising the benefits AI offers requires building and maintaining the public’s trust: citizens need to be confident AI is being developed and used in an ethical and trustworthy manner2 . AI poses considerable risks and challenges to society which have raised concerns about whether AI systems are worthy of trust. These concerns have been fuelled by high profile cases of AI use that were biased, discriminatory, manipulative, unlawful, or violated human rights. This survey is the first to take a deep dive into understanding citizen trust and expectations of AI use across multiple countries. To do this, we surveyed a nationally representative sample of 1,200+ citizens from the United States, Canada, Germany, the United Kingdom and Australia, respectively (total sample 6,054). We asked about AI systems in general, as well as AI systems in two domains – healthcare and human resources – where AI is rapidly being deployed and is likely to impact large numbers of citizens. Our findings provide important and timely research insights into citizens’ trust and attitudes towards AI. We draw on these insights to lay out an evidence-based pathway for strengthening trust and acceptance of AI systems, and discuss the implications for government, business and non-government organisations (NGOs). Below, we summarise the key findings. Most of these findings hold across all countries and therefore are reported in aggregate form. Significant country differences in the findings are highlighted.
Trust in artificial intelligence Trust is central to the acceptance of AI, and is influenced by four key drivers Our results confirm that trust strongly influences AI acceptance, and hence is important for the societal uptake of AI and realising its benefits. There are four key drivers that influence citizens’ trust in AI systems: 1) beliefs about the adequacy of current regulations and laws to make AI use safe, 2) the perceived impact of AI on jobs, 3) familiarity and understanding of AI, and 4) the perceived uncertain impact of AI on society. These drivers are important across all five countries. Of these drivers, the perceived adequacy of current regulations and laws is clearly the strongest. This highlights the importance of ensuring adequate regulatory and legal mechanisms are in place to protect people from the risks associated with AI use. Such regulation in turn supports citizen uptake and adoption. Citizens have low trust in AI systems but generally ‘accept’ or ‘tolerate’ AI Trust in AI is low across the five countries, with citizens generally wary or ambivalent about trusting AI systems. Only about a quarter (28%) of citizens are willing to trust AI systems in general. Two out of five citizens are unwilling to share their information or data with an AI system and a third are unwilling to trust the output of AI systems (e.g. a recommendation or decision). While many citizens are hesitant to trust AI systems, they generally accept (42%) or tolerate (28%) AI, but few approve (15%) or embrace (6%) it, and some outright reject AI (9%). Citizens’ trust and support of AI depends on the purpose of the AI system: the public is more trusting and supportive of AI use in healthcare (i.e. for aiding medical diagnosis and treatment), and less trusting and supportive of AI use in human resources (i.e. for aiding hiring and promotion decisions). Citizens also view the benefits of AI in healthcare as greater than the risks, whereas they view the risks of AI in human resources as greater than the benefits. However, regardless of the application, citizens are still wary with the majority unwilling or ambivalent about trusting AI in both healthcare (63%) and human resources (77%). Younger generations, notably Gen Z and Millennials, are generally more trusting and accepting of AI systems than older generations. In Germany and Australia, those with a university education are also more accepting of AI than those without a degree. Confidence in entities to develop, use and regulate AI varies across countries Citizens have the most confidence in their national universities and research institutions, as well as their defence organisations, to develop and use (71-77%) and regulate and govern AI (67-73%) in the best interest of the public. In contrast, citizens have less confidence in governments and commercial organisations to do this. 58% – 62% have confidence in commercial organisations and government to develop and use AI, and 54 – 58% have confidence in these entities to regulate and govern AI. This may be because most citizens believe commercial organisations (62%) and government (52%) innovate with AI for financial gain, rather than for societal benefit. Countries differ in their confidence of entities to use and govern AI. Americans are less confident in a broad range of entities to regulate and govern AI, compared to citizens in other countries. US and UK respondents are also less confident in their governments to develop and use AI in the public’s best interest compared to other countries, a finding that mirrors the lower trust these countries have in their governments more generally. In contrast, Australians are more confident in their research institutions and defence forces to develop, use and regulate AI. ©2021 The University of Queensland ©2021 KPMG, an Australian partnership and a member firm of the KPMG global organisation of independent member firms affiliated with KPMG International Limited, a private English company limited by guarantee. All rights reserved. The KPMG name and logo are trademarks used under license by the independent member firms of the KPMG global organisation. Liability limited by a scheme approved under Professional Standards Legislation. Trust in artificial intelligence 3 Citizens expect AI to be regulated with external, independent oversight The large majority of citizens (81%) expect AI to be regulated. While there are small country differences, there is general agreement (61-62%) that there should be a new, dedicated independent AI regulator, and that government and existing regulators should play a role in the regulation of AI systems. Co-regulation and involvement of industry that develop or use AI is also seen as desirable by the majority (54-59%). US respondents are less likely than respondents in other countries to report that government and existing regulators should regulate AI, and more likely to believe AI regulation is not required. In contrast, UK respondents are more likely to expect a new, dedicated, independent AI regulator than other countries. Current safeguards are insufficient given the uncertainty around AI Despite the strong expectations of AI regulation, most citizens (67%) across the five countries either disagree or are ambivalent that current regulations and laws are sufficient to make the use of AI safe. This powerfully highlights the importance of strengthening and communicating the regulatory and legal framework governing AI (including data privacy laws) across all surveyed countries. Most citizens (66-79%) in each country believe the impact of AI on society is uncertain and unpredictable. It is therefore not surprising that the large majority (96%) expect AI governance challenges to be carefully managed. The public view data challenges such as surveillance (61%), fake online content (60%), cyber-attacks (60%), and data privacy (59%) to be the most likely to impact large numbers of citizens within their country in the next 10 years. Half also viewed disease misdiagnosis as likely to impact society. Citizens expect organisations to uphold the principles of trustworthy AI Citizens in each country have very clear expectations of the principles and related practices they expect organisations deploying AI systems to uphold in order to be trusted. These principles mirror those proposed by the European Commission’s High Level Expert Group on AI. Almost all citizens (95%) expect AI systems to meet high standards of: – performance and accuracy – data privacy – security and governance – transparency and explainability – accountability – risk and impact mitigation – fairness – human oversight Most citizens (more than 57%) would be more willing to use AI systems if assurance mechanisms were in place, such as independent AI ethics reviews, AI ethics certifications, national standards for transparency, and AI codes of conduct. Organisations can directly build trust and consumer willingness to use AI systems by supporting and implementing these mechanisms. ©2021 The University of Queensland ©2021 KPMG, an Australian partnership and a member firm of the KPMG global organisation of independent member firms affiliated with KPMG International Limited, a private English company limited by guarantee. All rights reserved. The KPMG name and logo are trademarks used under license by the independent member firms of the KPMG global organisation. Liability limited by a scheme approved under Professional Standards Legislation. 4 Trust in artificial intelligence Citizens feel comfortable with some but not all uses of AI at work Only about one in five citizens (22%) believe AI will create more jobs than it will eliminate. Despite this, 45% of employed citizens report using AI in their work, but most use AI rarely or occasionally. Most citizens (70-76%) are comfortable with the use of AI at work for the purposes of task automation and augmentation. However, they are less comfortable with the use of AI for employeefocused activities, such as monitoring and evaluating employees, or in recruitment and selection. Citizens want to know more about AI but currently have low awareness and understanding of AI and its uses Most citizens (62%) have heard about AI. However, three out of every five citizens report a low understanding of AI, including how and when it is used in everyday applications. For example, even though 76% of citizens use social media, only 41% are aware social media uses AI. Men and the university-educated are more likely to be aware of AI and understand when it is being used. The good news is that most citizens across all countries (83%) want to know more about AI. Considered together, the results suggest there is a strong need and appetite for a public AI literacy program. A pathway to strengthen public trust in AI Collectively these survey insights provide an evidence-based pathway for building and maintaining the trust and acceptance of AI systems by citizens of western nations. As we discuss in detail in the concluding section, this pathway requires government and business to take action by: 1) living up to citizens’ expectations of trustworthy AI, 2) strengthening the regulatory framework for governing AI, and 3) enhancing AI literacy of the public and employees. The survey insights are relevant for informing AI policy and practice within business, government, and NGOs at the national level, as well as multinational and pan-governmental AI policy and practice (e.g. the Global Partnership on AI). Resources are available to support organisations to embed the principles and practices of trustworthy AI into their everyday operations, and put in place mechanisms that support stakeholder trust in their use of AI3 . Given the rapid investment and deployment of AI, it will be important to regularly re-examine public trust and expectations of AI systems as they evolve over time, to ensure AI use is aligned with and meeting societal expectations.
The benefits and promise of AI for society and business are undeniable. AI is helping people make better predictions and informed decisions, enabling innovation, productivity gains and improved efficiency, and lowering costs. It is helping protect physical and financial security and facilitating the global fight against COVID-19, to name just a few of its beneficial applications. The risks and challenges that AI poses for society are equally undeniable. These include the risk of codifying and reinforcing unfair biases, infringing on human rights such as privacy, spreading fake online content, technological unemployment and the dangers stemming from mass surveillance technologies, critical AI failures and autonomous weapons. These issues are causing public concern and raising questions about the trustworthiness and regulation of AI systems6 . The public’s trust in AI technologies is vital for continual acceptance. If AI systems do not prove to be worthy of trust, their widespread acceptance and adoption will be hindered, and the potentially vast societal and economic benefits will not be fully realised7 . Despite the central importance of trust, to date little is known about citizens’ trust in AI or what influences it across countries. Prior public attitude surveys8 have instead examined general acceptance and support. In 2020, we conducted the first deep dive survey examining Australians’ trust in AI systems (Lockey, Gillespie & Curtis, 2020). This report extends this deep dive on trust in AI by examining citizen perspectives across five nation states: the United States, Canada, Germany, the United Kingdom and Australia. This multi-country survey is designed to understand and quantify citizens’ trust in and support of AI, to benchmark these attitudes over time, as well as explore similarities and differences in trust and expectations across five western countries. Understanding similarities and differences across countries is important given AI systems are not bounded by physical borders and are rapidly being deployed and used across the globe. By taking this deep dive into the question of trust, this research provides a comprehensive and nuanced understanding of US, Canadian, German, UK and Australian citizens’ trust in AI systems. The research provides insights into the key drivers of trust, community expectations and confidence in the regulation of AI and management of societal challenges associated with AI. It also sheds light on citizens’ current understanding and awareness of AI, and the practices and principles citizens expect organisations to use to responsibly design, develop, deploy and govern AI in society and the workplace.