《大西洋理事會(Atlantic Council)：2024生產式AI時代背景下的老齡社會數字包容研究報告（英文版）（16頁）.pdf》由會員分享，可在線閱讀，更多相關《大西洋理事會(Atlantic Council)：2024生產式AI時代背景下的老齡社會數字包容研究報告（英文版）（16頁）.pdf（16頁珍藏版）》請在三個皮匠報告上搜索。

1、Generational AI:Digital inclusion for aging populationsAtlantic CouncilGEOTECH CENTERCAROLINE THOMPSONThis report has been made possible through the generous support of AARP.All views expressed in the report may not necessarily reflect the views of AARP.Throughout this process,we engaged in confiden

2、tial consultations with many well-known private and public organizations.These discussions were instrumental in shaping the con-tents of this report.Consequently,to maintain confidentiality,specific affiliations are not disclosed in the report.The Atlantic Council GeoTech Center is thankful for thes

3、e organizations insights.The author is grateful for the assistance of the following Geo-Tech Center staff:Raul Brens Jr.,Ellie Creasey,Coley Felt,Serena Lee,Brendan Oliss,Josh Orszag,Trisha Ray,and Lloyd Whitman.The mission of the GeoTech Center is to shape the future of technology and data to advan

4、ce the planet,people,prosperity,and peace.The center works to ensure emerging technologies enter use across the globe for public benefit,while identifying and mitigating potential geopolitical risks.Recent programs have focused on glob-ally promoting human-centered governance of artificial intellige

5、nce;bridging data divides;increasing the humanitarian-focused expansion of next-generation space-based communications;and enhancing the capacity of governments to use,regulate,and promote a range of emerging technologies.This report is written and published in accordance with the Atlantic Council Po

6、licy on Intellectual Independence.The authors are solely responsible for its analysis and recommendations.The Atlantic Council and its donors do not determine,nor do they necessarily endorse or advocate for,any of this reports conclusions.May 2024Cover image:KJ Pargeter via FreepikGenerational AI:Di

7、gital inclusion for aging populationsCAROLINE THOMPSONAtlantic CouncilGEOTECH CENTERTABLE OF CONTENTSINTRODUCTION 11.1 Background.11.2 Why focus on older adults?.1USE CASES AND UNIQUE CONSIDERATIONS FOR OLDER POPULATIONS 22.1 Use cases.22.1.1 Considerations:design.32.1.2 Considerations:development.3

8、2.1.3 Considerations:deployment.4EXISTING GAPS IN THE DIGITAL INCLUSION OF OLDER ADULTS IN THE AI LIFECYCLE 43.1 Gap 1a:incomplete or biased data on older adults.43.2 Gap 1b:lack of inclusion of older adults in AI design and development.53.2.1 Recommendation 1:forge data-inclusion and transparency s

9、tandards to integrate the experience of older adults in AI design and development.63.3 Gap 2:limited digital literacy and algorithmic awareness of older adults.63.3.1 Recommendation 2:empower user education and literacy for older adults,while ensuring proportional and appropriate modes of consent.73

10、.4 Gap 3:limited corrective action post-AI deployment.83.4.1 Recommendation 3:establish a standard of care through monitoring,evaluation,and impact assessments.8CONCLUSION 81ATLANTIC COUNCILGenerational AI:Digital inclusion for aging populationsINTRODUCTIONAs artificial intelligence(AI)applications

11、become ubiqui-tous in products and services,it is more important than ever to ensure that they are appropriately aligned for positive use and avoid exacerbating social exclusions for an aging population.Easily integrated into both existing and new processes,artificial intelligence is defined by the

12、Organisa-tion for Economic Co-operation and Development(OECD)as a system that analyzes input to generate outputs such as pre-dictions,content,recommendations,or decisions that can in-fluence physical or virtual environments.1 While the integration of AI will affect nearly every population,it will do

13、 so in various ways depending on factors unique to each industry and pop-ulation segment.Based on discussions with leaders in equi-ty,AI,and aging,and additional research,the following brief outlines the unique considerations for older adults within the AI lifecycle,barriers to digital inclusion tha

14、t older adults expe-rience regarding AI and suggested near-and long-term solu-tions to advance digital inclusion and mitigate biases against older adults,while supporting practical AI innovation,AI policy,and healthy aging.1.1 BackgroundAI applications amplify the environment in which they are cre-a

15、ted,reflecting the preferences or biases of their development teams or datasets,which may compound data,algorithmic,or societal biases at scale as AI is integrated into everyday uses of emerging technologies.This issue is particularly salient for aging populationsa sizeable,but often neglected,popul

16、a-tion segment of users of emerging technologies such as AI.Following the release of OpenAIs ChatGPT tool in November 2022,the widespread integration of AI accelerated a digital world where technology began to shape society as much as society shaped technology,and those lacking access or in-clusion

17、in AI discourse could be left behind.There are many positive use cases for AI with aging populations,such as im-proving community health,individual wellness,health equity,and supporting productivity and social relations.2 However,without proper guidance to align AI design and deployment with respons

18、ible,inclusive,and equitable practices,these 1“OECD AI Principles Overview,”Organisation for Economic Co-Operation and Development,November 29,2023,https:/oecd.ai/en/ai-principles.2“Request for Information(RFI)on Strengthening Community Health Through Technology;Correction,”Federal Register,April 2,

19、2022,https:/www.federalregister.gov/docu-ments/2022/02/04/2022-02289/request-for-information-rfi-on-strengthening-community-health-through-technology-correction.3 Philip Meylan,et al.,“Digital Inclusion for All:Ensuring Access for Older Adults in the Digital Age,”Foreign Policy Analytics(2023),2,htt

20、ps:/ Kieran Walsh,Thomas Scharf,and Norah Keating,“Social Exclusion of Older Persons:A Scoping Review and Conceptual Framework,”European Journal of Ageing 14,1(2016),8198,https:/ Equitable Healthy Aging in Low-and Middle-Income Countries:The Aging Readiness&Competitiveness Report 4.0,”AARP,2023,34,h

21、ttps:/www.aarpinternational.org/file%20library/arc/aging-readiness-competitiveness-report.doi.10.26419-2fint.00053.001.pdf.6 Charlene H.Chu,et al.,“Digital Ageism:Challenges and Opportunities in Artificial Intelligence for Older Adults,”Gerontologist 62,7(2022),948,https:/ Diaz,et al.,“Addressing Ag

22、e-Related Bias in Sentiment Analysis,”Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems,April 2018,https:/dl.acm.org/doi/10.1145/3173574.3173986.7 Ittay Mannheim,et al.,“Inclusion of Older Adults in the Research and Design of Digital Technology,”International Journal of En

23、vironmental Research and Public Health 16,19(2019),3718,https:/ Alexander Peine and Louis Neven,“The Co-Constitution of Ageing and Technologya Model and Agenda,”Cambridge University Press Ageing and Society 41,12(2020),2846,https:/www.cambridge.org/core/journals/ageing-and-society/article/coconstitu

24、tion-of-ageing-and-technology-a-model-and-agenda/0778C39BAE292B6E-B409E91B04440124.technologies can also lead to further ageism,confusion,or general misuse.In particular,age and its intersection with other dimensions of accessincluding income,race,language,and genderdramatically influence an individ

25、uals ability to fully access,benefit from,and contribute to the digital world.3 When older adults undergo social exclusion within the digital world,this has widespread effects on their quality of life and the equity of an aging society.4 Intertwined with additional forms of exclu-sion such as econom

26、ic,institutional,socio-political,territorial,and symbolic,social exclusion can be amplified by digital sys-tems developed or deployed without inclusive practices.5 A common misperception exists that older adults are ill-suited for the digital realm and incapable of using technology,and this misperce

27、ption of older adults as technologically illiterate hinders the full consideration of their needs as users during the design and deployment of AI technologies.Researchers can be influenced by ageist sentiment,potentially affecting AI products and services that might manifest harms of allocation(dist

28、ribution of resources and opportunities)and representa-tion(how groups are represented and perceived by society).6 In reality,older adults can have rich knowledge and technical and occupational skills that will be lost if digital tools are inac-cessible.Involving older adults in development and disc

29、ourse around AI can reduce ageist stigma in design,research,and society,and can promote older adults agency and unique voice in society.7Aging and technology influence one another,creating a co-con-stitution that elevates the need to ensure that the development and implementation of AI technologies

30、are done in ways that empower older populations rather than disengage or disregard them.8 Guidance and policies that include and engage older adults in AI development and deployment can foster broader inclusion,as the demographic cuts across various protected statuses and minority identities.Empower

31、ing the inclusion of older adults supports them in acting as agents of enhancing more comprehensive inclusion across AI.This change is neces-sary to ensure responsible and equitable AI for all.1.2 Why focus on older adults?The global population is rapidly aging and that trend is set to continue,with

32、 an expected 1.4 billion people aged sixty 2 2ATLANTIC COUNCILGenerational AI:Digital inclusion for aging populationsand older by 2030a 56-percent increase from 2017.9 Rec-ognizing this,the United Nations designated 20212030 as the Decade of Healthy Ageing to secure healthy and digni-fied futures fo

33、r older adults.10 By 2034 in the United States,Americans over sixty-five are projected to outnumber the pop-ulation of youth under eighteen;by 2060,they will represent more than one-quarter of the US population.11The impending demographic shifts,both within the United States and across the world,wil

34、l have significant market im-plications.Older adults already qualify as the wealthiest age group in the world,but big tech companies often overlook them despite their purchasing power.12 However,their spend-ing on new technologies is surging.US adults over fifty were estimated to spend$912 on tech o

35、n average in 2023,up 130 percent from 2019.13 By 2030,their total tech expenditure could reach$108 billion,with their economic contributions in the United States potentially tripling to$26.8 trillion by 2050.14 As companies race to launch AI-enabled products to serve consumers,they are incentivized

36、by first-mover advantage and profits to create generalized products.These incentives and practices can not only overlook populations willing to en-gage in technological movements,like older adults,but can also create a misalignment between the outcomes of these products.Processes and incentives that

37、 encourage business-es to receive funding and develop people-first models will help advance technological equity.It is critical to include older adults in technological-develop-ment discourse,particularly with regard to AI,to achieve eq-uitable digital development.With a positive view of technol-ogy

38、s societal impact,older adults are not only adopters but also advocates for their peers.Whether through their impact on medical,emotional,or social health,such as active com-munity engagement and integration to stave isolation and exclusion,AI policies must consider the unique needs of the aging pop

39、ulation,ensuring digital inclusion and engagement across all facets of life,at home or work,to positively impact the lives of older adults and promote healthy aging.Older adults continue to live healthier lives,remaining engaged in independent economic and informal community activities,while activel

40、y contributing to many different societal roles.15 Despite their impact,many technology and AI policies and frameworks are not explicitly considering older adults when 9“Ageing and Health,”World Health Organization,October 1,2022,https:/www.who.int/news-room/fact-sheets/detail/ageing-and-health.10 J

41、ustyna Stypinska,“AI Ageism:A Critical Roadmap for Studying Age Discrimination and Exclusion in Digitalized Societies,”AI&Society 38,2(2022),665677,https:/ Technologies to Support an Aging Population,”Committee on Technology of the National Science&Technology Council,2019,https:/trumpwhitehouse.arch

42、ives.gov/wp-content/uploads/2019/03/Emerging-Tech-to-Support-Aging-2019.pdf;Luke Rogers and Kristie Wilders,“Shift in Working-Age Population Relative to Older and Younger Americans,”United States Census Bureau,June 25,2020,https:/www.census.gov/library/stories/2020/06/working-age-population-not-keep

43、ing-pace-with-growth-in-older-americans.html.12 Meylan,et al.,“Digital Inclusion for All.”13 Brittne Kakulla,“2023 Tech Trends and the 50-Plus,”AARP,January 5,2023,https:/www.aarp.org/pri/topics/technology/internet-media-devices/2023-technology-trends-old-er-adults/.14“Expanding Digital Inclusion fo

44、r Aging Populations,”Foreign Policy Analytics,2022,https:/ Rights of Older Persons:The Data Gap,”United Nations General Assembly Human Rights Council,2020,https:/undocs.org/A/HRC/45/14.16 Mannheim,et al.,“Inclusion of Older Adults in the Research and Design of Digital Technology.”17 Diaz,et al.,“Add

45、ressing Age-Related Bias in Sentiment Analysis.”18 Katherine Wild,et al.,“Unobtrusive In-Home Monitoring of Cognitive and Physical Health:Reactions and Perceptions of Older Adults,”Journal of Applied Gerontology 27,2(2008),181200,https:/ specific demographics,including the March 2024 United Nations

46、AI Resolution and the draft UN Global Digital Compact.To ensure digital inclusion across the en-tire AI lifecycle,comprehensive AI design,development,and deployment policy should consider the aging populations unique circumstances and capabilities.USE CASES AND UNIQUE CONSIDERATIONS FOR OLDER POPULA

47、TIONS2.1 Use casesMany AI-enabled technologies currently on the market and utilized by older adults focus on health or chronic disease management,instead of more universal applications for pro-ductivity or entertainment.16 Rather than supplementing a normal phase of life,many of these existing healt

48、h-focused AI technologies for older adults treat aging as a problem that technology can solve.17 Like the general population,older adults may interact with AI directly and indirectly.When considering direct interactions,most are within the category of ambient assisted-living(AAL)health technologies,

49、such as wearables that may monitor physical decline or disruptions in patterns,including falls or heart perturbations.18 Beyond wearables,older adults may di-rectly interact with health technology such as cognitive orthot-ics(automated reminder systems),home technologies such as voice-assistant tech

50、nology,or digital companion tools.Indi-rectly,older adults may interact with AI through AI-generated content and algorithms utilized by service providers,such as healthcare,financial,or housing services.When products apply to all people for all purposes,their effec-tiveness wanes,and their resulting

51、 generalized outputs may eventually negatively affect subsets of the population.De-signing,developing,and deploying AI-enabled technologies specifically designed with,rather than for,populations can al-leviate this misalignment.Many of these principles constitute best practices for user design acros

52、s many technologies and experiences,but it is important to emphasize how these best practices can be used to better integrate and support older 3ATLANTIC COUNCILGenerational AI:Digital inclusion for aging populationsadults in AI use cases,as their circumstances are often over-looked.As older adults

53、interact with AI-enabled technologies,there are considerations unique to their age that affect inter-actions with the following stages of the AI lifecycle:design,development,and deployment.2.1.1 Considerations:designWhen designing AI technologies inclusive of the needs of old-er adults,four consider

54、ations should be reviewed,including:function,form,and compatibility;sensory abilities;cognitive abilities;and the overall diversity of older adults.Designers should consider creating technologies that support healthy aging and complement human interaction,which are essential for maintaining relation

55、ships and community.Func-tionality is critical;it is important to address the real needs of older adults,rather than just the most publicized ones.Addi-tionally,designers must also consider cognitive,sensory,and physical-ability variations within this group.Technologies that are cumbersome or ineffe

56、ctive are likely to be disregarded by older adults,but this should not be mistaken for a lack of in-terest or ability.Designing interoperable tools and processes that are easily compatible with existing habits and lifestyles alleviates a significant barrier to technology adoption.Further-more,while

57、age may be a factor in sensory decline,it does not automatically imply cognitive decline.Recognizing this diver-sity through design would empower older adults by placing agency back in their hands.Beyond function,AI developers need to view older adults as a diverse group with intersecting characteri

58、stics,rather than a uniform group defined solely by age.This approach is essen-tial to mitigate ageism and social bias within AI design.Social bias often arises from multiple dimensions and can dispropor-tionately affect older adults,especially those with disabilities,those living alone,those with l

59、ess education,and those who are non-native English speakers.19 By identifying and address-ing these intersecting inequalities from the design stage,AI developers can“drive multidimensional change and maximize impact”of their solutions.20 This perspective shift is critical to creating a more inclusiv

60、e and effective approach that caters to older adults needs.2.1.2 Considerations:developmentDuring the development of AI technologies,five consider-ations for older adults that should be top of mind include:19 Meylan,et al.,“Digital Inclusion for All,”6.20 Ibid.,7.21 “Aging Connected:Exposing the Hid

61、den Connectivity Crisis for Older Adults,”Humana Foundation and Older Adults Technology Services,2021,6,https:/agingconnected.org/report/.22 “Community Connected Health Stakeholder Engagement Summary Report,”White House Office of Science and Technology Policy,May 2022,https:/www.whitehouse.gov/wp-co

62、ntent/uploads/2022/05/05-2022-Summary-of-CCH-Stakeholder-Engagement.pdf.23 Barbara Barbosa Neves,et al.,“Artificial Intelligence in Long-Term Care:Technological Promise,Aging Anxieties,and Sociotechnical Ageism,”Journal of Applied Gerontology 42,6(2023),1278,https:/ Gerontechnology refers to the int

63、ersection of technology and aging.internet access;variety of platforms;interoperability;diversity of end users;and validation processes.Internet access is a crucial determinant of how older adults may engage with AI-enabled technologies,influencing the when,where,and how of their interactions.This a

64、ccess var-ies significantly across different environments,such as home,work,or community settings,and can depend on the device used,whether it is a computer,tablet,or smartphone.Nota-bly,many existing AI-enabled services designed for older adultssuch as Medicare Part D signup,telehealth,and fi-nanci

65、al servicesare optimized for larger screens and not mobile phones,potentially excluding those primarily using smartphones for internet access.21 To promote digital inclusion among older adults,it is essential to consider the range of internet and device access levels and the platforms they use.Moreo

66、ver,in the context of older adults using AI,the actual end users of these systems can be quite diverse,extend-ing beyond older adults to include younger adults,informal and formal caregivers,clinicians,family members,and other groups across the continuum of care.These additional users facilitate acc

67、ess for the approximately 1.3 million Americans living in assisted-living facilities and eight hundred thousand in nursing homes.For providers and caregivers who assist older adults,adopting digital technologies hinges on seamless inte-gration with existing technology,workflows,and systems.22 In-ter

68、operability is critical to lower barriers to technology use and support integration with current practices and lifestyles.The positive impact of AI-enabled tools and processes can only go so far if adopters are overburdened by excessive inter-faces and devices.Applications for AI in long-term care o

69、ffer benefits beyond just monitoring,including social connections,detection,and clinical uses such as pain assessment and medication adherence.They can automate tasks,increasing staff interaction time with residents and enhancing overall pa-tient care.23 Recognizing the diverse roles of end users,fr

70、om older individuals to long-term care(LTC)staff and clinicians,is vital throughout gerontechnologys design,development,and deployment.24 Additionally,consistent and framework-aligned output-validation processes can enable developers to mit-igate any biases present throughout development.These proce

71、sses will help development teams course correct within testing and sandbox environments before reaching consum-ers.This approach helps mitigate bias,ensure technology flu-ency,and safeguard user agency.4 4ATLANTIC COUNCILGenerational AI:Digital inclusion for aging populations2.1.3 Considerations:dep

72、loymentIn deploying AI technologies,there are three key consider-ations to ensure digital inclusion for older adults,including:enhancing digital literacy;fostering algorithmic awareness and transparency;and ensuring informed consent or appropriate,proportional alternatives.A 2021 survey by Foreign P

73、olicy Analytics revealed that 54 percent of US adults over fifty wanted better digital skills,yet 37 percent lacked confidence in using technology.25 This gap in confidence and skill highlights the need for improved digital literacythe understanding of digital tools,including AI.De-ploying tools tha

74、t are already compatible with the lives of old-er adults,whether through interoperable interfaces or through the backend of existing systems,helps relieve the burden on individuals to improve digital literacy.By enabling tools that are more advanced and impactful versions of existing pro-cesses,inno

75、vative solutions can empower older adults,care-givers,and healthcare providers without burdening them with the responsibility of understanding complex technical pro-cesses or an excessive number of systems.Equally important to literacy and interoperability are algorithmic awareness and transparency,

76、which involve being informed about when and how AI systems are utilized,particularly understanding how an individuals data may be used,which can better empower old-er adults to interact with these systems.A lack of information about an algorithm or a gap in awareness around whether an algorithm is i

77、n use can lead to what researchers call an algo-rithmic divide.26 This gap can prevent effective engagement with technology,exacerbating digital exclusion.Informed consent extends beyond just acknowledging,through disclosure,the use of AI.It includes understanding how the system uses ones data,its i

78、mpact on outputs,and ongoing data utilization.Fundamental to informed consent is algorithmic awareness,which indicates communication in-forming or knowledge of algorithms in use.Careful consider-ation must be taken to utilize consent practices proportional to the type,use,and output of algorithms.Fo

79、r instance,there are particular use cases where the likelihood of harm from not using data,or delayed use of data,outweighs the likelihood of harm from using data,such as privacy concerns for a patient,user,or population.At a minimum,disclosure of the use of AI informs awareness while utilizing adva

80、nced tools for positive outcomes that mitigate the harms of delayed use or inaction.Insufficient algorithmic awareness can lead to lower or inade-quate use of applications,diminishing their representation in data sets that train and validate AI.From a market standpoint,lower algorithmic awareness ca

81、n also make the older popula-tion appear as a less relevant audience to private companies developing and deploying AI.27 This can reinforce a cycle of 25 Meylan,et al.,“Digital Inclusion for All,”5.26 Stypinska,“AI Ageism.”27“Ageism in Artificial Intelligence for Health,”World Health Organization,Fe

82、bruary 9,2022,https:/www.who.int/publications/i/item/9789240040793.28 Charlene H.Chu,et al.,“Age-Related Bias and Artificial Intelligence:A Scoping Review,”Nature Humanities and Social Sciences Communications 10,510(2023),1,https:/ exclusion,especially because AI tools are predomi-nantly developed a

83、nd implemented by private companies that might overlook the older demographic.If tech companies do not adequately consider older adults,this can create a neg-ative feedback loop,where their exclusion from AI develop-ment leads to further digital exclusion.EXISTING GAPS IN THE DIGITAL INCLUSION OF OL

84、DER ADULTS IN THE AI LIFECYCLEWhen considering policy or programming to support the digital inclusion of older adults in AI,four significant barriers should be addressed.They are:incomplete or biased data on older adults;lack of inclusion of older adults in AI design,development,and post-deployment

85、feedback;limited digital literacy and algorithmic awareness of older adults;and adaptive monitoring and evaluation.These process gaps can perpetuate digital ageism,where age-ism is reflected in the“design,development,and implemen-tation of AI systems and technologies and its resultant data.”28 Under

86、standing these gaps in digital equity helps inform poten-tial near-and long-term solutions to support the digital inclusion of older adults throughout the entire lifecycle of a technology.Each gap can be addressed by a recommendation that cen-ters on empowering older adults,while increasing the effi

87、ca-cy of AI algorithms and applications.Fostering inclusion can alleviate the misconception that older adults are technopho-bic when,in reality,many are willing and able to engage in interfacing with new technologies.Most recommendations can be implemented across private industry,civil-society or-ga

88、nizations,universities,and local governments,which do not face the same bureaucratic lag of the federal government.To address each gap,priorities suggested for the multistakehold-er field of AI development,deployment,and governance are:forging data-inclusion and transparency standards;empowering use

89、r education and literacy for older adults,while ensuring proportional and appropriate modes of consent;and establishing a standard of care through monitoring,evalu-ation,and impact assessments.5ATLANTIC COUNCILGenerational AI:Digital inclusion for aging populations3.1 Gap 1a:incomplete or biased dat

90、a on older adultsThe challenges in AI technology development for older adults stem from two significant gaps in data handling:the scarcity of representative data and the oversimplification of existing data into broad,uniform categories.The issue starts at the data-or-igin and training phase of AI sy

91、stems,where biases can be introduced due to ineffective or incomplete data labeling and annotation.These practices often reflect ageist assumptions,treating older adults as a uniform group without considering their diverse abilities and needs.29 Such biases in data sets become amplified when applied

92、 in AI technologies.One of the critical issues is the lack of comprehensive data on older adults,particularly when training generative AI models.Often,data collected on older adults focus on those living in-dependently,neglecting those in long-term care facilities or multigenerational households.Thi

93、s oversight leads to a sig-nificant gap,as living arrangements do not necessarily cor-relate with the ability or inclination to utilize new technologies.As a result,AI applications developed with such data fail to align the needs and realities of a substantial portion of the older population.Moreove

94、r,generative AI models often suffer from sampling bias,as sources like social media,where old-er adults are less represented,are commonly used for data collection.30 The result creates invisible users,as older adults considerations and experiences are not considered in tech-nological design,developm

95、ent,and deployment,making their interests and values negligible to processes.31The problem extends to the qualitative aspects of data used in AI.Research on supervised learning models(models with labeled data)has found a polarizing effect in age coding,with adjectives such as young,younger,and young

96、est often scored more positively than adjectives such as old,older,and oldest.32 The semantic bias in sentiment analysis,crucial in sectors like marketing and politics,reflects cultural and histor-ical biases encoded in the data.Furthermore,in data collec-tion,older adults are frequently lumped into

97、 broad categories,such as those fifty-five and older or sixty-five and older.This approach fails to capture the nuanced experiences at differ-ent stages of aging,unlike the more specific age categoriza-tions used for younger adults.Addressing these issues is particularly urgent in healthcare technol

98、ogies,a sector heavily utilized by older adults.33 The prevailing view of older adult data as a minority or edge case in AI development leads to technologies that do not ade-29 Ibid.,12.30 Ibid.,12;“Human Rights of Older Persons.”31 Chu,et al.,“Digital Ageism,”949.32 Diaz,et al.,“Addressing Age-Rela

99、ted Bias in Sentiment Analysis,”1.33“Ageism in Artificial Intelligence for Health.”34 Kakulla,“2023 Tech Trends and the 50-Plus.”35 Bjrn Fischer,Alexander Peine,and Britt stlund,“The Importance of User Involvement:A Systematic Review of Involving Older Users in Technology Design,”Gerontologist 60,7(

100、2019),e514,https:/ Can AI Positively Impact Aging Populations?”private roundtable,Atlantic Council,Washington,DC,August 6,2023.37 Fischer,et al.,“The Importance of User Involvement,”516.38 Joe Kita,“Workplace Age Discrimination Still Flourishes in America,”AARP,December 30,2019,https:/www.aarp.org/w

101、ork/age-discrimination/still-thrives-in-america/.quately serve this significant user group.The public feels this exclusion,with a majority of older Americans feeling that mod-ern technology is not designed with them in mind.34 Policies promoting the inclusion of diverse and detailed data from older

102、adults are essential for developing equitable AI applica-tions,especially in healthcare,where the impact of exclusion can be most profound.3.2 Gap 1b:lack of inclusion of older adults in AI design and developmentThe development of AI-enabled technologies often over-looks the specific needs of divers

103、e user groups,particularly older adults.While universal design is a common approach,it is crucial to recognize the value of actively involving various end users,including older adults,in the design process.This involvement can range from low-level engagement,such as completing questionnaires or cons

104、ultations,to more intensive forms of participation,such as participating as an equal partner in design processes.35 Such inclusion can significantly shape the development of AI technologies,ensuring they meet the diverse needs of all potential users.36 Often,older users are only engaged in initial i

105、nterviews or post-design evaluations,limiting their impact on the development process.Participatory design,in which users actively contribute to the creation process,reflects democratic principles and can be particularly empowering for older adults.This approach aligns the design with the users actu

106、al needs and can significantly bridge digital divides.There are several motivators for involv-ing older adults in the design process.First,it fosters a deep-er understanding and learning environment,which is critical to overcoming digital exclusion.Second,designing with the users needs in mind resul

107、ts in higher-quality products more likely to be accepted and used by the target population,such as ensuring interoperability with existing processes to lower barriers to adoption.Last,and perhaps most important,partic-ipatory design returns agency to older adults,allowing them to contribute as exper

108、ts of their own experiences and needs.37 However,a significant challenge in achieving this inclusive design approach is the lack of diversity in technology-de-velopment teams.The tech industry,particularly in Silicon Valley,skews toward younger demographics,with the medi-an age significantly lower t

109、han the national workforce aver-age.38 This age gap can perpetuate ageist biases and create uneven power dynamics,often excluding the perspectives and needs of older adults.The continuous exclusion of older adults from the design process reflects and reinforces so-6 6ATLANTIC COUNCILGenerational AI:

110、Digital inclusion for aging populationscietal ageism.This can lead to internalized ageism among older adults,reducing their willingness to engage with new technologies.Such a pattern of non-use further exacerbates their exclusion,creating a feedback loop of digital marginal-ization.39 Addressing thi

111、s gap requires a conscious effort to include older adults in technology design and development,ensuring their unique perspectives and needs are adequate-ly represented and addressed.3.2.1 Recommendation 1:forge data-inclusion and transparency standardsto integrate the experience of older adults in A

112、I design and development.The AI Ethics Guidelines Global Inventory,which covers 146 English documents from various sectors such as government,private,civil society,and international organizations,reveals that only a minority(23.3 percent)recognize ageism as a bias in AI.40 This oversight is signific

113、ant,given the diverse and growing population of older adults interacting with AI technol-ogies.The United States has made some progress,with the nonbinding September 2022 White House Blueprint for an AI Bill of Rights acknowledging age and older adults as classes deserving protection from algorithmi

114、c discrimination.This rec-ognition is a crucial step toward addressing the unique needs of older Americans in AI equity.However,there is a need for policies that actively mitigate ageist systemic risks,such as design and discourse exclusion,to avoid the high probability and impact risks associated w

115、ith ageism and social exclusion.Improving practices to include representative data on older adults in AI design and development is crucial to mitigate bias from the outset and lay a foundation for equity throughout AIs lifecycle.Transparency in AI,particularly in disclosing which demographics are in

116、cluded or excluded in datasets,is es-sential for understanding an AI tools purpose and limitations.Engaging older adults in design can not only create more in-novative products,but also increase adoption of innovative tools by considering interoperability with existing practices.Additionally,fosteri

117、ng intergenerational interactions can be a powerful tool in reducing ageism.41 Policies encouraging such interactions between people of different generationswheth-er in community,educational,or employment spacescan help alleviate intergroup prejudice and stereotypes,benefit-ing older adults and youn

118、ger developers.Ensuring inclusion and understanding are applied to educating current and new generations of AI professionals can operationalize policy that encourages inclusive practices.39 Hanna Kttl,et al.,“But at the Age of 85?Forget It!:Internalized Ageism,a Barrier to Technology Use,”Journal of

119、 Aging Studies 59(2021),100971,2,https:/ Chu,et al.,“Digital Ageism,”951.41“Ageism in Artificial Intelligence for Health.”42“Global Digital Compact:Zero Draft,”United Nations Tech Envoy,April 1,2024,7,https:/www.un.org/techenvoy/sites/www.un.org.techenvoy/files/Global_Digital_Compact_Zero_Draft.pdf.

120、43“Seizing the Opportunities of Safe,Secure and Trustworthy Artificial Intelligence Systems for Sustainable Development,”United Nations General Assembly,2024,6,https:/documents.un.org/doc/undoc/ltd/n24/065/92/pdf/n2406592.pdf?token=sWIanlBbjgTKiQbDMJ&fe=true.44 Monica Anderson and Andrew Perrin,“Tec

121、h Adoption Climbs Among Older Adults,”Pew Research Center,May 17,2017,https:/www.pewresearch.org/internet/2017/05/17/tech-adoption-climbs-among-older-adults/.45 Brittne Kakulla,“Older Adults Embrace Tech but Are Skeptical of AI:2024 Tech Trends and Adults 50-Plus,”AARP,December 19,2023,https:/www.aa

122、rp.org/pri/topics/technol-ogy/internet-media-devices/2024-technology-trends-older-adults/.While frameworks and regulations take time,steps to create inclusive discourse can be implemented almost immediately.Many mainstream debates and conversations regarding fair-ness and inclusivity in AI fail to c

123、onsider age as a critical di-mension.Elevating the necessity to acknowledge and discuss how age intersects with AI,and how AI intersects with age,is crucial to encouraging tech equity within both technology companies and policymaking bodies.Although the recent executive order on AI from the White Ho

124、use,“EO 1411 Safe,Secure,and Trustworthy Development and Use of AI,”men-tions bias nine times,it falls short of explicitly considering the unique needs of different age groups.Additionally,the April 2024 first draft of the United Nations Global Digital Compact notably only includes children as a spe

125、cific age group when citing the need for age diversity in designing and developing technologies.42 Similarly,the United Nations first AI resolu-tion,A/78/L.49,adopted in March 2024,only explicitly states gender,race,and disability as notable gaps within data and connectivity divides.43 Policymakers

126、might consider including age,and specifically referencing older adults as a specific age group,alongside protected statuses such as race and sex,to ensure a more inclusive approach to AI development and reg-ulation.3.3 Gap 2:limited digital literacy and algorithmic awareness of older adultsA 2017 Pe

127、w Research Center survey highlighted that nearly half of the older adult respondents needed assistance to set up or learn to use new electronic devices.44 The finding under-scores the challenges faced by older adults who have adapt-ed to technological advancements later in life,particularly as digit

128、al literacy becomes more complex with the advent of ar-tificial intelligence.The lack of understanding and confidence in navigating these technologies can lead to an accessibility gap,affecting independence,productivity,and engagement with AI systems.Furthermore,insufficient digital literacy can exp

129、ose older adults to risks like misuse of AI systems or so-cial engineering,such as misaligned uses of generative AI by malicious actors including misinformation campaigns,cyber schemes,voice spoofing,and phishing.Older adults perceptions of AI-enabled tools vary,with 60 percent of Americans aged fif

130、ty and above surveyed by AARP saying that they are undecided about generative AIs impact.45 Some may view AI-enabled tools as intrusive,while others see them as empowering aids for independent living,supporting their private lives with extra safety assurance through moni-7ATLANTIC COUNCILGenerationa

131、l AI:Digital inclusion for aging populationstoring and alert tools.46 This dichotomy underscores the im-portance of proportional consent,which involves clear and comprehensive agreements on data collection,usage,and implementation as they relate to the purported use of the data.Proportional consent

132、is based on the practice of pro-portionality in data protection,where“proportionality requires that advantages due to limiting the right are not outweighed by the disadvantages to exercise the right.”47 Proportional con-sent is especially critical to the equitable deployment and use of many AI appli

133、cations,and is particularly salient to the appli-cations most likely to be used by older adults.In addition to the plethora of AI-enabled healthcare appli-cations used by or for older adults,research in 2016 found that older adults are more likely to access personal informa-tion-management applicati

134、ons than other age groups.48 These applications,which often include sensitive and personal infor-mation about a user,such as frequent contacts and locations,require clear communication about data accessibility,usage,and privacy.Older adults,more likely to use personal man-agement applications,need a

135、ssurance about who can access their data and how their data will be used,in order to build trust and ensure they remain in control of their information.Addressing this gap requires improving the digital literacy of older adults and ensuring transparent and user-centric design in AI technologies.3.3.

136、1 Recommendation 2:empower user education and literacy for older adults,while ensuring proportional and appropriate modes of consent.The National Artificial Intelligence Advisory Committee(NA-IAC)recommends a national AI literacy campaign to equip the public with the resources and knowledge needed t

137、o under-stand,embrace,trust,or adapt to an AI-infused world.To succeed,these AI literacy initiatives should involve collabo-ration between various groups beyond the federal government,including industry,nongovernmental organizations,and aca-demia.These programs can be most effective with older popu-

138、lations when conducted by technologically adept peers in trust-ed community organizations,including churches and libraries.49 Beyond similar-age peers,intergenerational support can also help build trust and confidence in utilization and understand-ing.A critical component of digital literacy is cont

139、extualizing AI and algorithms to better understand that their outputs are not ground truth,and they will never be free of bias.Most outputs will approximate the data available within a models dataset and,therefore,should always be subject to human questioning or objection.Advances in digital literac

140、y work in tandem with consumer trust.As older adults learn how to interact with AI sys-tems and what the limitations of those systems are,they will be 46 Peine and Neven,“The Co-Constitution of Ageing and Technology,”2852.47“Necessity and Proportionality,”European Data Protection Supervisor,last vis

141、ited March 18,2024,https:/www.edps.europa.eu/data-protection/our-work/subjects/necessi-ty-proportionality_en#:text=In%20the%20context%20of%20fundamental,disadvantages%20to%20exercise%20the%20right.48 Andrea Rosales and Mireia Fernndez-Ardvol,“Beyond WhatsApp:Older People and Smartphones,”Romanian Jo

142、urnal of Communication and Public Relations 18,1(2016),27,https:/journalofcommunication.ro/index.php/journalofcommunication/article/view/200.49 Meylan,et al.,“Digital Inclusion for All,”5;Kttl,et al.,“But at the Age of 85?Forget It!”6.better prepared to make informed decisions on when to utilize or

143、rely on outputs from AI-enabled tools.Increasing AI literacy is crucial to addressing digital and algo-rithmic divides.However,any efforts to do so must also con-sider the unique needs of older adults and the intertwined re-lationship between consent and AI literacy.AI literacy should increase aware

144、ness and critical understanding and should empower individuals to understand their options when choos-ing whether to engage with AI tools.Creating a foundation of trust via algorithmic awareness enables methods of propor-tional consent that balance data rights while mitigating possi-ble harms of com

145、plicated and costly consent models,which can delay the positive output of algorithmic processes or even lead to exacerbated harms.Especially in cases of AI in health-care,time is of the essence.Modernizing consent practices to include proportionality protects the agency of individuals,while also all

146、owing technologies to make positive impacts.The non-binding Blueprint for an AI Bill of Rights notes the necessity of informed consent and the ability to withdraw consent.Withdrawn consent will remain a key component of modernized consent practices for AI.Within healthcare,multi-ple organizations ar

147、e working toward principles and codes of conduct for health AI centered around patient consent,includ-Digital Literacy for CaregiversThere is a window of opportunity to provide digital literacy not only to older adults but also to those who may support them.Further devel-opment support for caregiver

148、s,such as virtual learning networks or application programming interface(API)integrations to minimize disrup-tions to user-facing systems,can help caregiv-ers successfully integrate AI tools for good in their work with older adults.Ageist tendencies can also appear in geron-tology professionals unde

149、rstanding,imple-mentation,application,and uptake of AI tech-nologies.New systems can reinforce these inequities and lead to further exclusion.Comprehensive efforts to elevate digital litera-cy and algorithmic awareness among caregiv-ers can mitigate potential negative power dy-namics when older adul

150、ts are not the primary end users of an AI technology.8 8ATLANTIC COUNCILGenerational AI:Digital inclusion for aging populationsing the National Academies of Medicine,the Coalition for AI Health,the Health AI Partnership,and the OECD AI Principles and Classification Framework.3.4 Gap 3:limited correc

151、tive action post-AI deploymentAI systems are highly adaptable and evolve rapidly,requir-ing policy solutions that continuously safeguard the interests of specific populations,such as older adults,in varying and evolving contexts.As the World Health Organization empha-sized in 2022,“if governments,in

152、tergovernmental agencies,publicprivate partnerships and nongovernmental organiza-tions do not exercise sufficient oversight of AI technologies,”older populations cannot trust that AI technologies are de-signed or deployed appropriately on their behalf.50 Without this oversight,there is a risk of ero

153、ding public trust,which is an essential function of healthy democracies.The current lack of thorough monitoring and evaluation of AI systems,mainly those used by older adults,risks exacerbat-ing exclusion from design and development and issues that can be scaled by AI applications including fraud,da

154、ta misuse,and misinformation.The misalignment of AI systems with their intended purpose can render them ineffective or harmful.Im-plementing AI solutions without fully understanding the users needs can be counterproductive.Regular assessments and adjustments are essential,especially in the context o

155、f AI-en-abled healthcare technologies that are frequently targeted at older adults.Ongoing evaluations by international and civil-so-ciety organizations invested in technological and aging equi-tysuch as the World Health Organization,the United Nations through the new Global Digital Compact or UN AI

156、 Resolution,AARP,the Center for AI and Digital Policy,OECD,the World Economic Forum,Data and Society,or new initiativescan en-sure these technologies remain functional,safe,and aligned with their intended purpose.3.4.1 Recommendation 3:establish a standard of care through monitoring,evaluation,and i

157、mpact assessments.The United Nations Educational,Scientific and Cultural Or-ganization(UNESCO)Recommendation on the Global Ethics of AI,adopted in 2021,emphasized the need for continuous monitoring,evaluation,and impact assessments of AI applica-tions to align these applications with ethical standar

158、ds.In the long term,continuous monitoring and evaluation will be critical to safeguard individuals and communities against exacerba-tion of harm through AI applications.This alignment can be made possible by building processes and systems for impact assessments that will not only track and quantify

159、the impact of AI systems,but also provide thresholds and processes for when and how to take corrective action on AI systems.The current liability structure of AI systems is insufficient.De-spite widespread daily interaction with these systems,there is a lack of accountability for ensuring fair and j

160、ust outcomes 50“Ageism in Artificial Intelligence for Health.”beyond frameworks.A duty of care should be established for automated systems collecting data,ensuring the systems are not trusted without question and are subject to human review.Establishing a standard of care through industry best pract

161、ices of monitoring,evaluation,and impact assessments will enable technologists to correct and rectify programs that become misaligned or otherwise cause harmful results.Policymakers should enforce these standards by requiring corrective mon-itoring and action to provide a layer of consumer protectio

162、n that will be especially necessary for groups that face more bias than others,including older adults.CONCLUSIONThe recent developments in guidelines,frameworks,and agreements signify a positive shift toward enabling digital in-clusion for older populations.These developments are crucial to safeguar

163、d against biases inherent in AI-enabled technolo-gies,biases that can significantly impact older adults through-out the various stages of the AI lifecycle.The path forward demands not just the inclusion of older adults in AI,but also their empowerment.As AI products and services become in-tertwined

164、with daily life,advocating for the rights and needs of the aging population becomes more critical.This approach will pave the way for an equitable landscape where older Americans are not merely passive recipients,but active con-tributors and beneficiaries of the AI revolution.This progress,however,i

165、s just the beginning.Realizing an in-clusive digital environment requires continuous efforts in both the short and long terms.These efforts should focus on am-plifying the voices of older adults in AI discourse,ensuring their perspectives and experiences shape the development and application of AI t

166、echnologies.By doing so,we can create a future where AI not only serves the diverse needs of all age groups,but also respects and values the unique contributions of older adults.Ultimately,this will lead to a more just and eq-uitable society where the benefits of AI innovation are shared by everyone

167、,regardless of age.9ATLANTIC COUNCILGenerational AI:Digital inclusion for aging populationsAUTHOR BIOCaroline Thompson is an assistant director at the GeoTech Center,where she manages projects and research at the in-tersection of geopolitics and emerging technologies,with a particular focus on artif

168、icial intelligence.Thompson joined the GeoTech Center upon completing the Atlantic Council Young Global Professionals Program.Before joining the At-lantic Council,she interned with the US Department of State at the embassy in Budapest and worked in higher education and software.Thompson is proficien

169、t in French,and her areas of interest include innovation,data bias,national security,and artificial intelligence.Thompson received her BA in international relations and his-tory from Tulane University and her MA in national security,focusing on critical and emerging technologies,from Amer-ican Unive

170、rsity.During her studies,Thompson was a fellow for the Center for Security,Innovation,and New Technology,a research assistant for digital policy,and she completed re-search on artificial intelligence and national security at Cam-bridge University.1010ATLANTIC COUNCILGenerational AI:Digital inclusion

171、 for aging populations11ATLANTIC COUNCILGenerational AI:Digital inclusion for aging populations*Executive Committee Members List as of April 24,2024 CHAIRMAN?EXECUTIVE CHAIRMAN EMERITUS?PRESIDENT AND CEO?EXECUTIVE VICE CHAIRS?VICE CHAIRS?TREASURER?DIRECTORS?Jarosaw Grzesiak?HONORARY DIRECTORS?The At

172、lantic Council is a nonpartisan organization that promotes constructive US leadership and engagement in international affairs based on the central role of the Atlanticcommunity in meeting todays global challenges.2024 The Atlantic Council of the United States.All rightsreserved.No part of this publi

173、cation may be reproduced ortransmitted in any form or by any means without permissionin writing from the Atlantic Council,except in the case ofbrief quotations in news articles,critical articles,or reviews.Please direct inquiries to:Atlantic Council1030 15th Street,NW,12th Floor,Washington,DC 20005(202)463-7226,www.AtlanticCouncil.org