PaperlayerView exact revision on arXiv

Source-linked AI summary

How does HCI Understand Human Autonomy and Agency?

Dan Bennett, Oussama Metatla, Anne Roudaut, Elisa Mekler

arXiv:2301.12490v2cs.HC

TL;DR

HCI research uses autonomy and agency in diverse, often ambiguous ways, while technologies increasingly influence behaviour and values. This paper reviews HCI work on the concepts, finding broad but under-specified meanings and proposing dimensions for clarifying and connecting future research.

  • Problem

    HCI uses autonomy and agency to describe diverse phenomena, leaving unclear how researchers understand the concepts and whether those understandings are compatible as technologies influence behaviour and values.

    Source: §1 · p. 1

  • Method

    The authors review 161 papers on human autonomy and agency, using an inductively developed thematic coding rubric to analyze the corpus.

    Source: §3.2 · p. 3, §3.3.1 · p. 4

  • Results

    HCI researchers broadly value autonomy and agency, but the review finds diverse understandings and treats the concepts as vague umbrella terms often used interchangeably.

    Source: §1 · p. 2, §5 · p. 9, §6 · p. 12

  • Takeaways & Limitations

    Four aspects can help clarify approaches and situate prior work, while future research can relate local understandings and coordinate perspectives across HCI communities.

    Source: §1 · p. 2, §6 · p. 12

    The review excludes adjacent constructs such as empowerment, efficacy, and competence, leaving their relationship to autonomy and agency for future work.

    Source: §5.4 · p. 12

Abstract

from arXiv · show

Human agency and autonomy have always been fundamental concepts in HCI. New developments, including ubiquitous AI and the growing integration of technologies into our lives, make these issues ever pressing, as technologies increase their ability to influence our behaviours and values. However, in HCI understandings of autonomy and agency remain ambiguous. Both concepts are used to describe a wide range of phenomena pertaining to sense-of-control, material independence, and identity. It is unclear to what degree these understandings are compatible, and how they support the development of research programs and practical interventions. We address this by reviewing 30 years of HCI research on autonomy and agency to identify current understandings, open issues, and future directions. From this analysis, we identify ethical issues, and outline key themes to guide future work. We also articulate avenues for advancing clarity and specificity around these concepts, and for coordinating integrative work across different HCI communities.

1 INTRODUCTION

Autonomy and agency are longstanding HCI concerns made more pressing by technologies that influence behaviour and values. The review finds broad, often interchangeable meanings and proposes ways to clarify them and coordinate future work.

1 INTRODUCTION

  • Persuasive technologies, intelligent systems, and tighter technological integration into people’s lives and bodies heighten the relevance of autonomy and agency.These developments raise implications for human autonomy and agency.Source: §1 · p. 1
  • HCI accounts leave unclear how autonomy and agency are understood across communities and whether those understandings are compatible.The concepts are sometimes treated as synonymous and sometimes as distinct but entangled.Source: §1 · p. 1
  • A systematic literature review presents findings from 161 papers spanning 32 years of research, beginning in 1990.Source: §1 · p. 1
  • The review finds that autonomy and agency operate as vague umbrella constructs, are often used interchangeably, and encompass diverse meanings.Few works specifically define the concepts or examine their relationships to expected outcomes.Source: §1 · p. 2
  • Four aspects—self-causality and identity, experience and material expression, timescales, and independence versus interdependence—can help distinguish approaches and guide future work.The paper also outlines ways to clarify the concepts and their relationships to valued outcomes, and to coordinate work across HCI communities.Source: §1 · p. 2

2 BACKGROUND

The paper situates HCI’s diverse uses of autonomy and agency within a history of entangled concepts and argues that existing accounts do not yet provide a broad, integrated view. It frames its review as a way to clarify distinctions and connect work across communities.

2 BACKGROUND

  • The review focuses on human autonomy and agency, discussing autonomous technologies and nonhuman agency only when they bear on those human concerns.Source: §2 · p. 2
  • Autonomy is associated with self-governance and agency with self-causality, but the concepts are often entangled or treated as near-synonymous.Their distinction is not always easy to draw in everyday life.Source: §2 · p. 2
  • Accounts link agency and autonomy to causal influence, values, identity, and social context, across both extended decisions and moment-to-moment experiences.The cited work describes influence from personality and social factors on objective correlates of low-level agency.Source: §2 · p. 2

2.1 Agency and Autonomy in HCI

  • HCI has addressed autonomy and agency through design principles, frameworks for well-being, user experience, and research on closely coupled human-technology systems.Examples include Value-Sensitive Design, METUX, and Human-Computer Integration.Source: §2.1 · p. 2
  • HCI work ranges from momentary control to gaming and dementia, but shared language obscures how these approaches relate across behavioural scales.Understanding their interrelations matters for considering possible effects of technology delegation at sensorimotor timescales on decision-making and everyday life.Source: §2.1 · pp. 2–3
  • METUX offers life spheres for organizing design thinking, but does not focus specifically on autonomy or connect its account to the breadth of prior HCI work.The review aims to survey that breadth and identify how its approaches relate and might coordinate.Source: §2.1 · p. 3

3 REVIEW METHOD

The authors review HCI publications spanning 32 years, selecting papers about human autonomy or agency and analyzing them with a thematically developed coding rubric. Full-corpus coding was calibrated through independent coding and discussion.

3 REVIEW METHOD

  • The review aims to overview how HCI researchers have addressed and understood agency and autonomy across 32 years of research.Source: §3 · p. 3

3.1 Source Selection

  • A search of five HCI venues in the ACM Digital Library for “agency” and “autonomy” in keywords and abstracts returned 207 full papers.The search was conducted on May 21, 2022.Source: §3.1 · p. 3

3.2 Selection of papers for inclusion in the review

  • After screening for human autonomy or agency and excluding 46 papers focused on nonhuman issues without discussion of human concerns, 161 papers remained.Source: §3.2 · p. 3

3.3 Coding Procedure

  • The authors developed an inductive coding rubric through thematic analysis of 77 papers, grouping 466 initial codes into seven categories and validating decisions with coauthors.The rubric was informed by background reading and intended to capture explicit and implicit approaches.Source: §3.3.1 · pp. 3–4

3.3.1 Developing the Coding Rubric.

  • The first author coded all 161 papers using the rubric; two authors independently coded a random subset of 13 papers and resolved differences through discussion.Only minor rubric revisions were needed, clarifying definitions of timescales.Source: §3.3.2 · p. 4

4 RESULTS

The results section introduces the analysis sequence, moving from subject matter and attributed values to authors’ understandings and the aspects found across the corpus.

4 RESULTS

  • The analysis proceeds from subject matter and attributed benefits to authors’ understandings, then to aspects of agency and autonomy; Table 1 summarizes the categories.Source: §4 · p. 4

4.1 Subject matter

The reviewed work spans many HCI domains, with autonomy and agency often framed differently across them: as material support in accessibility-related work and as felt control in entertainment.

4.1 Subject matter

  • Ageing and accessibility formed the largest group (n=33), often addressing personhood, social ties, and independent living amid health or material constraints.Similar material concerns appeared in work on parenting, children, and HCI4D.Source: §4.1 · p. 4
  • Gaming, social media, and entertainment formed the next largest group (n=32), usually emphasizing how free, independent, self-governing, or in control people felt.Source: §4.1 · p. 4
  • Other subjects ranged from work (n=17) to input methods (n=6).Source: §4.1 · p. 4

4.2 Value and Outcomes

HCI papers valued autonomy and agency intrinsically and for varied benefits to users and organizations, but often left their value implicit and sometimes described competing interests.

4.2 Value and Outcomes

  • In 57 papers, the value of autonomy and agency was implicit or unclear, despite their being widely regarded as desirable qualities.Source: §4.2 · p. 4
  • Twenty-one papers treated autonomy or agency as intrinsically valuable, including as a basic human need, a right, an ethical value, or part of human dignity.Source: §4.2.1 · p. 4

4.2.2 Beneficial Outcomes of Agency and Autonomy for the User.

  • Most papers (n=90) located their value in user benefits, including well-being, life satisfaction, overcoming barriers, health, and positive development.Source: §4.2.2 · p. 4
  • Other user benefits included better user experience, control, enjoyment, meaning, learning, privacy, security, and practical outcomes.Some papers linked autonomy and agency to safety, security, or privacy tensions, particularly for children or people with diminished capacities.Source: §4.2.2 · pp. 4–5

4.2.3 Beneficial Outcomes of Agency and Autonomy for Other Stake-

  • Twenty-three papers described benefits accruing beyond individual users, including to social groups, organizations, or companies.Source: §4.2.3 · p. 5
  • Organizational benefits did not always align with users’ values or interests; some accounts paired organizational gains with user burdens or uncertain benefits.Examples included added worker stress and labour, or a user “voice” that legitimized existing organizational goals.Source: §4.2.3 · p. 5

4.3 How Did Authors Articulate Understandings of Agency and Autonomy?

Authors articulated autonomy and agency unevenly: explicit definitions were limited, meanings were often implied or unclear, and few papers engaged established theories.

4.3 How Did Authors Articulate Understandings of Agency and Autonomy?

  • Only one third of papers (n=58) explicitly defined autonomy or agency, and some definitions were unspecific or tautological.Source: §4.3 · p. 6
  • In 47 papers, meanings were implied through associative statements rather than explicit definitions.Source: §4.3 · p. 6
  • In 55 papers, meanings were unclear or inconsistent, including shifts between decision making and task execution or between network-level and artifact agency.Source: §4.3 · p. 6
  • Some papers identified autonomy or agency as a significant theme in their abstract or introduction but did not use the terms again.Source: §4.3 · p. 6
  • Very few papers explicitly engaged with existing theories; Self-Determination Theory was most frequently mentioned (n=17), alongside frameworks for particular research foci.Cognitive-science concepts such as intentional binding appeared in work on immediate control over actions.Source: §4.3 · p. 6

4.4 Aspects of Autonomy and Agency

The review organizes HCI accounts of autonomy and agency around self-causality and identity, material and experiential aspects, time-scales, and independence or interdependence. Across these dimensions, the corpus shows that the concepts cover varied and sometimes overlapping phenomena.

4.4.1 Self-causality and Identity.

  • 159 papers addressed causality and 54 identity; execution appeared in 122, while 99 covered both execution and decision and 36 focused mainly on choice.Some papers distinguished decision from execution, and loss of executional involvement was associated with diminished agency, flexibility, or adaptation.Source: §4.4.1 · pp. 6–7
  • Self-congruence appeared in 41 papers and self-construction in 24; conflicting goals can make it unclear whether an outcome aligns with a user’s values.Self-congruence concerns alignment with values and goals regardless of involvement in decisions or execution, whereas self-construction concerns how values and identity develop.Source: §4.4.1 · p. 7
  • Self-construction accounts considered how technologies may shape identity, values, and goals, including risks of reinforcing dependency and examples of self-definition or reflective self-change.Examples include assistive technology, children’s video-viewing self-control apps, and self-tracking for chronic health conditions.Source: §4.4.1 · p. 7

4.4.2 Material and Experiential.

  • 80 papers focused exclusively on material aspects, 46 exclusively on experience, and 33 on both; minimal interactions could alter felt agency without causal influence, though the reviewed cases did not show divergence between the aspects.Material accounts include abilities and constraints on acting or pursuing preferences; experiential accounts include the felt sense of agency or autonomy.Source: §4.4.2 · p. 7

4.4.3 Time-scales.

  • The corpus distinguishes micro-interactions, interaction episodes, and longer spans of life; approaches differed across these scales, though few papers explicitly compared them or addressed multiple scales.Some work identified tensions, trade-offs, and synergies between immediate and long-term autonomy.Source: §4.4.3 · pp. 7–8
  • 72 papers studied interaction episodes, 35 wider-life scales, and 13 micro-interactions; six micro-interaction papers used intentional binding, whose relation to conscious experience remained unclear.Episode-focused papers all addressed self-causality, while identity-related aspects appeared in 13; 17 of the wider-life papers addressed identity.Source: §4.4.3 · p. 8
  • 41 papers addressed both episodes and wider life; some described accepting short-term restrictions for longer-term agency, while everyday execution could support self-definition over time.Examples include blocking digital distractions to pursue personal goals and accepting website vetoes to retain longer-term freedom to browse independently.Source: §4.4.3 · p. 8

4.4.4 Independence or Interdependence?

  • 50 papers centered independence and 54 interdependence; social networks and living contexts could support autonomy, while dependence could also be experienced as undermining it.The corpus includes accounts that translate autonomy into independence without further distinction, as well as critiques of accounts that underrecognize interdependence.Source: §4.4.4 · pp. 8–9

5 DISCUSSION

The review finds broad agreement that autonomy and agency matter, but their meanings and relationships remain diverse and weakly coordinated across HCI. It argues for clarifying local aspects while relating them across communities and outcomes.

5 DISCUSSION

  • Researchers value autonomy and agency, yet treat them as vague umbrella concepts whose distinctions are mostly implicit and rarely connected across HCI domains.The review found few cases of work building on findings beyond its immediate domain, despite commonalities across contexts.Source: §5 · p. 9

5.1 Clarifying Agency and Autonomy as Boundary Objects

  • Making autonomy and agency effective boundary objects requires locally refined meanings and integrative coordination with wider understandings.The paper argues for both local attention to relevant aspects and work relating those aspects to one another.Source: §5.1 · pp. 9–10

5.1.1 Articulate Locally Relevant Aspects of Agency and Autonomy.

  • Researchers should explicitly define locally relevant aspects; the review’s four distinctions offer flexible, non-exhaustive categories grounded in HCI work.The four aspects concern self-causality and identity, experiential and material aspects, timescales, and independence and interdependence.Source: §5.1.1 · p. 10

5.1.2 Investigate Relationships Between Different Aspects.

  • Studies rarely connect autonomy and agency across timescales or other dimensions; future work can relate episodes in everyday life to choices, social context, and UX.The review also suggests relating temporal-binding measures to predictions from Self-Determination Theory.Source: §5.1.2 · p. 10

5.2 Clarifying how Aspects of Agency and Autonomy Relate to Outcomes

  • Links to UX and high-value outcomes remain underspecified: temporal binding–UX associations were asserted before direct tests found the relationship unclear, while long-term outcomes were not tracked.The corpus contained no studies comparing high-value outcomes before and after sessions or observing them over the long term.Source: §5.2 · p. 11

5.3 Ethical Challenges

  • Vague claims about autonomy support can obscure who benefits and mask manipulation; claims should match plausible outcomes to the aspect and timescale studied.The review found examples where users gained limited local control or only the experience of agency while accepting additional burdens or restrictions.Source: §5.3 · pp. 11–12
  • Self-change interventions may restrict short-term autonomy, so researchers should weigh what is restricted or supported over time against users’ goals and values.The review calls for reasoning about the relevant aspects, timescales, consequences, and warrant for interventions.Source: §5.3 · p. 12

5.4 Limitations and Open Questions

  • The review excludes adjacent constructs such as empowerment, efficacy, and competence, and its 32-year scope treats agency and autonomy at a relatively high level.The authors suggest that domain-specific reviews could clarify understandings in particular communities and contexts.Source: §5.4 · p. 12

6 CONCLUSION

The review finds that HCI treats autonomy and agency as important but ambiguous umbrella concepts, often used interchangeably. It identifies four aspects to clarify them and calls for more explicit definitions, specific outcome links, and integrative work across communities.

6 CONCLUSION

  • Across 161 papers spanning 32 years, autonomy and agency appear as broad, often interchangeable concepts with diverse meanings and theoretical underpinnings.This ambiguity makes it difficult to treat the concepts as a whole or relate different understandings to outcomes and prior work.Source: §6 · p. 12
  • The review identifies four aspects for clarifying HCI understandings: self-causality and identity, experiential and material expression, timescales, and independence and interdependence.These aspects can help researchers locate relevant prior work and situate their own research.Source: §6 · p. 12
  • The authors call for explicit definitions, use of the four aspects, more specific links to outcomes, and integrative work connecting aspects and HCI communities.The aim is to clarify local understandings while coordinating them with wider community understandings.Source: §6 · p. 13
Loading 2301.12490v2…