After hundreds of applications, you finally land an interview for your dream job. To your shock, the first round isn’t with a person, it’s with an algorithm that scans your responses to pre-recorded questions. It scores you poorly for a few “ums,” your discomfort with the AI format, and your confusion about the process—shutting down your opportunity before a human ever sees your application. 

Scenarios like this raise pressing questions about the ethics of automated decision-making (ADM), a domain within AI ethics that examines the moral and societal implications of delegating choices to algorithmic systems.  In this context, ADM encompasses algorithmic systems, ranging from rule-based to self-learning, that collect and analyze data to generate outputs guiding or substituting human decision-making. ADM now impacts nearly every aspect of society, representing a monumental shift in decision-making environments that previously relied on human experts.¹ In its application spanning from approving credit cards and guiding autonomous vehicles to suggesting medical diagnoses and predicting mental health disorders, the ethics of ADM must be discussed to fully understand both the benefits and risks.^{2, 3, 4} The ethics of ADM judges how algorithms guide choices with respect to human agency and oversight. 

When we think about ADM, technical frameworks like decision trees or neural networks might come to mind, but we must take a broader perspective to understand the ethics behind it.¹ Key features like autonomy, efficiency, and scalability for complicated decision-making scenarios factor into our evaluation, as well as the reality that many ADM models are now AI-based decision-makers that replace or supplement human experts. With automated systems taking on levels of agency previously reserved for humans, the behavioral design and application of ADMs need to embody humanist values and ethics in its interventions.⁵

What primary ethical challenges come with ADM systems?

In our discussion on the ethics of ADM, we must consider what the key ethical concerns are when employing these systems. These challenges highlight how automated systems not only make technical decisions, but also shape trust and accountability in human terms:

While we’ve captured some of the core ethical challenges with ADM, they are just some of the key areas where guidelines are needed. As a versatile solution to potential issues, researchers suggest that ethics-based auditing may be a feasible way to support the governance of organizations that are using ADM to make high-stakes choices.¹ Ironing out ethical regulations, ensuring ADM systems abide by them, and earning the trust of stakeholders is no small feat. To simplify this solution, we can visualize the dynamics as a series of relationships and paths of information exchange between organizations and human agents:¹ 

Ethical scrutiny for ADM, with ethical AI solutions

At the crux of ADM, there are promises of autonomy, efficiency, and scalability for solving complex problems. The paradox of these qualities is that they may magnify harms if we don’t analyze the ethics of automated processes. It is important to recognize possible trade-offs between efficiency and fairness, as well as consistency and trust. For these reasons, ethical scrutiny of ADM and AI systems is necessary to ensure they strengthen ethics in institutional and organizational decision-making.¹¹
If we pursue ethical scrutiny for ADM systems, then we need frameworks that can sufficiently capture the risks. Two key models that can guide how we design, use, and update ADM are responsible AI and ethical AI, which both pave the path forward for ADM systems that prioritize high ethical standards.^{12, 13, 14} While these models overlap in their goal of upholding principles like transparency and accountability, they differ in their specifics. Let’s take a closer look at these ethics-oriented AI models while considering the promises and risks of ADM in practice:

Responsible AI and ethical AI are complementary lenses that help us push ADM toward safer and fairer outcomes. In principle, they uphold governance and human rights alike without compromising human agency. With the rise of generative AI, the future of ADM demands powerful safeguards to manage novel risks and mandate responsible use.

Ethics-Based Auditing: The systematic evaluation of AI and automated decision-making systems to ensure they align with ethical standards and societal values.¹ This may include key ethical features like transparency and accountability, or verifying regulation through legal frameworks.

Algorithmic Bias: Systematic and repeatable errors in AI outputs that result in unfair treatment of certain individuals or groups, often reflecting underlying societal or data-driven prejudices. For example, an algorithm that assumes that someone in a poorer neighborhood should get a lower credit score.

Algorithmic Opacity: The difficulty of understanding or explaining how complex models, like neural networks and machine learning, reach their conclusions.¹⁶ With high opacity, there is less transparency on the ethical risks of ADM.

Proxy Variables: Data points that unintentionally stand in for sensitive attributes, such as zip code acting as a proxy for race or income.¹⁷ In the context of ADM, proxy variables can result in discriminatory outcomes even when protected characteristics are not explicitly included in the dataset despite seeming “neutral.”

Algorithmic Fairness: The design and use of algorithms to mitigate systemic discrimination that ensures equitable treatment in automated decision-making.¹⁸ Within ADM, fairness is essential for maintaining public trust and preventing algorithms from reinforcing the inequalities that they aim to resolve.

Responsible AI: The design, deployment, and governance of AI systems in a manner that is transparent and accountable.¹² In practice, this focuses on compliance and risk management relative to procedures to ensure AI is managed responsibly while producing positive social impact.

Ethical AI: AI systems that are developed in accordance with ethical principles for transparency and fairness.⁵ This may include respect for human rights and justice to ensure that decisions made by ADM respect human dignity in alignment with broader social norms. 

The ethics of automated decision-making trace back to the origins of algorithmic systems in government in the 1960s. Early versions of automation held the contemporary promise of efficiency, while surfacing immediate concerns of fairness in contexts like determining credit scores and welfare eligibility, where proxy variables often played a role.¹⁹ Scholars in computer science and philosophy considered how computational models may reinforce biases, with specific concerns about prejudice in recidivism assessments and parole recommendations.¹⁹ Although formal conversations on ADM ethics were limited before the 2000s, seeds of accountability, responsibility, and the balance between efficiency and fairness were planted decades prior.²⁰

In the 1990s, algorithms moved into fields like healthcare, finance, and policing, which made the ethics of ADM hard to ignore. In 2000, computer scientist Latanya Sweeney demonstrated that algorithms may re-identify individuals in anonymous datasets, which implied the potential risk of algorithmic harm for certain groups compared to others.²¹ As a pioneer of data privacy and a frontrunner of algorithmic fairness, her work marked some of the first empirical research that showed the potential for algorithmic harm.²² Far ahead of the curve, Sweeney recognized our daily experience of algorithmic bias before it had a name. She shifted the ADM ethics debate from theory to reality, and from a tool once labeled as objective and neutral to a system that risks cycling large-scale inequality. 

The 2010s brought the intersection of automated decision-making and ethics under a magnifying glass as it reached public awareness. Thinkers like Cathy O’Neil took addressing algorithmic harm a step further: in her book Weapons of Math Destruction, she argued that algorithms have the tendency to mask discrimination with a veil of objectivity in hiring and education.²³ These types of critiques resonated as machine learning expanded into daily life, sparking urgent calls for transparency and accountability as “mega-tech” companies gained influence while central governments offered little guidance.²⁴ ADM ethics moved into mainstream discourse, leading policymakers, journalists, and civil society to become concerned about what happens when algorithmic power remains unchecked.

Looking to the late 2010s and early 2020s, ethical debates around ADM have been shaped by figures like Timnit Gebru, a computer scientist known for her groundbreaking work on algorithmic bias and AI ethics.²⁵ Gebru’s research on facial recognition highlighted systemic harms against marginalized communities, reinforcing the need for structural governance frameworks. Responsible AI and ethical AI models are now emerging in parallel to fill the gap of guidance for organizations as they blend upholding compliance, governance, and human rights values.²⁶ ADM ethics are no longer a niche concern but a global imperative at the intersection of technology and society.

In the past year, a major milestone came with the European Union’s AI Act, the world’s first comprehensive legal framework for artificial intelligence. Passed in 2024, it categorizes ADM systems by risk level and sets strict requirements for transparency, accountability, and human oversight.²⁷ This marks a historic shift: ethical concerns about ADM are now codified into law, signaling the maturation of the field from academic debate to enforceable international governance. What began as abstract theoretical warnings about ADM now appears in the rulings of courthouses, practices of technology companies, and human-centered approaches of behavioral science.²⁸

Latanya Sweeney
An American computer scientist and privacy expert recognized for pioneering research on algorithmic discrimination.²² She demonstrated how automated systems can unintentionally harm individuals and perpetuate inequities.

Cathy O’Neil
An American data scientist and author best known for Weapons of Math Destruction.²³ She highlighted how large-scale algorithms can reinforce inequality and social harm, emphasizing the societal consequences of opaque automated systems.

Timnit Gebru
An Ethiopian-American AI researcher and co-founder of the Black in AI initiative, who was fired from Google for raising issues of discrimination at work.²⁹ Her work focuses on algorithmic bias, ethical AI practices, and the broader societal impacts of machine learning on marginalized communities.

When ethical principles help us navigate automated decision-making, there are potential benefits beyond risk management. From enabling sustainable decisions to promoting well-being and fostering trust, these outcomes show how ADM can serve people and society responsibly.

Sustainable and scalable decision-making across public domains

Sustaining human decision-making can be difficult to maintain over time across domains. A primary benefit of ADM is its ability to scale decisions to the societal level, far beyond what individuals or even groups of humans could manage on their own. This efficiency supports the aspects of work that either strain our cognitive load capacities or involve repetitive, mundane efforts. For instance, in healthcare, ADM is expected to have an immediate effect in analyzing data and diagnoses—two common sources of both cognitive overload and tedium. In the future, it doesn’t feel hard to imagine having an AI medical “professional” with the necessary knowledge for a treatment decision.¹¹

Another context that has seen the effective use of ADM is policing, where algorithm-based surveillance and individual profiling allow law enforcement to analyze crimes for better public safety.³⁰ Paired with AI, these ADM systems improve operational capacity as well as the ability to investigate crimes, driving the transition from post-crime to pre-crime practices for risk mitigation of future criminal activity.³¹ From an ethical perspective, however, vigilance is required, as these same ADM models can make prejudicial choices related to racial profiling, privacy invasion, and the potential to reinforce systemic inequality.^{32, 33}

Maximizing well-being, minimizing harm

One way to understand the societal impacts of ADM ethics is from a utilitarian perspective, where there is an emphasis on creating the most happiness and well-being while mitigating harmful social impact. This raises further-reaching ethical questions, particularly in areas where ADM may entirely replace human decision-making. In a time of self-driving cars and AI companions, we need to ask what it means for an ADM system to act ethically, and further consider who takes responsibility when ADM slips up, to avoid undermining the key ethical features of transparency and accountability.³⁴

All of this can sound overly philosophical without tangible examples. One place where we may assume the imperative of maximizing well-being and minimizing harm is in risk assessment contexts like credit scoring.³⁵ When ADM is used to evaluate risk, the goal is to expand access to fair credit while reducing harmful bias. Without strong governance and auditing, however, these systems risk reproducing unfair outcomes at scale, showing how abstract ethical debates translate into real impacts on people’s lives.

When to rely on ADM and AI, and when to pause

Questions that arise surrounding ADM ethics are how and when we should trust automated processes for our choices, particularly when AI becomes the decision agent. When comparing automated agents to humans, one useful lens is how they handle uncertainty. A notable shift with ChatGPT-5 is that instead of fabricating an answer, it more openly admits “I don’t know” compared to earlier versions.³⁶ While this makes amends for embarrassing past mistakes and false generative AI facts, research with about 400 participants shows that when LLMs express uncertainty, our confidence in them decreases.³⁷
Other literature dives into how personal characteristics have an effect on the perception of ADM through AI across domains like media, public health, and judicial settings.³⁸ Using a scenario-based survey experiment with nearly 1000 participants, the authors found mixed opinions on what people think about fairness and the usefulness of ADM across society. While individual traits played a role in these perceptions, several participants rated decisions made by AI as good as or even better than those made by humans for specific decisions. This research reiterates that the promise of ADM is fulfilled when we can harness it for well-being without erasing fairness and accountability, with trust as a natural extension of these features.

The ethical debate on automated decision-making has several key criticisms. Some important ones to hone in on are the seesaw of bias vs. objectivity, transparency and the right to challenge automated decisions, and the necessity of human oversight.

Bias with a veil of objectivity 

ADM systems are often said to be neutral and scientific in contrast to flawed human judgment, yet they are prone to inheriting human prejudices via the data they are trained on. By appearing objective, these systems risk disguising discrimination as impartial judgment. As one long-form piece from Harvard explains, when algorithms influence decisions in fields like parole, hiring, or policing, they essentialize social biases with “a kind of scientific credibility” that makes them harder to contest.¹¹ The ethical difficulty lies in recognizing that efficiency and scale by no means guarantee fairness.

As bias and discrimination are major areas of ethical concern when considering the impacts of AI on society, real-world consequences demonstrate how damaging the illusion of objectivity of ADM can be. In the U.S., a widely used risk assessment tool for parole decisions was found to overestimate the likelihood of reoffending for black defendants, reinforcing racial disparities in the justice system.¹¹ Similar issues arise in automated hiring platforms, where word choices or educational background can unfairly disadvantage women and minority applicants. These cases highlight the risk of bias being legitimized by algorithms, raising urgent questions about how to audit and adjust ADM before they push inequality further.

Transparency and the right to challenge automated decisions 

One of the most pressing controversies is whether people can meaningfully contest ADM outcomes that affect their lives. The UK GDPR’s Article 22 codifies a right to human review, reflecting a broader demand for systems that are explainable and open to challenge.³⁹ Without transparency, individuals face opaque decisions that deny them jobs, credit, or services, with little clarity about why the system ruled against them. This lack of recourse undermines not only fairness but also ongoing trust in ADM systems.

Practical examples illustrate the stakes of gaps in transparency. Consider our earlier example of job applicants screened out by resume-sorting algorithms, who may never know whether biased keyword filters or irrelevant proxies like zip code triggered their rejection. In financial services, individuals can be denied loans without any explanation of the factors that led to their risk score. These black-box processes erode accountability, making it essential that ADM systems provide clear, auditable decision trails and accessible appeal mechanisms.

Gaps in human oversight and governance 

Even when ethical risks are known, effective oversight of ADM remains elusive. The UK’s Ethics, Transparency and Accountability Framework for Automated Decision-Making outlines governance best practices, but uptake is uneven and regulators often lack the technical depth to enforce them.40 Scholars argue that without systematic auditing and organizational-level governance, the harms of ADM cannot be contained at the algorithmic level alone.⁵ The challenge is not simply designing fair systems, but embedding accountability across institutions.

The gap between theory and practice has already produced tangible harm. In the Netherlands, the government deployed an ADM system called SyRI to flag welfare fraud, but the opaque model disproportionately targeted low-income and immigrant communities.^{41, 42} This result led to public outrage, as The Hague District Court made a ruling that the system violated human rights—including the right to privacy, failing to consider the public interest of locating welfare fraud. This case shows how inadequate oversight can permit ADM and AI alike to amplify systemic inequities, underscoring the need for governance that is transparent, enforceable, and accountable at scale.

Privacy and ADM: Navigating the black box

AI-based automated decision-making (ADM) is becoming central to operations in government and business, promising faster, more efficient processes. In Canada, federal agencies have launched hundreds of ADM projects, from triaging disability pension applications to shortlisting candidates to promote diversity in hiring.⁴³ These systems are often designed to assist human decision-makers rather than replace them, in a collaborative fashion. Yet, they raise pressing ethical questions. Large datasets may include sensitive personal information, and organizations may find it impractical to obtain consent for every use. The balance between efficiency and protecting individual privacy remains a critical challenge.

Beyond consent, ADM introduces risks around profiling and opaque decision-making. Systems that evaluate employment, financial, or government services may unintentionally discriminate, sometimes without individuals even knowing. For example, a Chinese tutoring company settled a U.S. lawsuit after its AI hiring software rejected older applicants.⁴³ ADM can affect access to opportunities, legal rights, and social mobility, raising complex questions about fairness and accountability. Experts highlight the need for explainable AI and clear processes for human review to prevent harm and maintain trust.

The regulatory landscape is evolving but remains fragmented. The European Union’s Artificial Intelligence Act imposes a tiered, risk-based approach, requiring key ethical elements like transparency and human oversight as necessary components for high-risk AI systems.⁴³ Canada participates in international initiatives, such as the Council of Europe’s Framework Convention on AI, but domestic legislation is limited. Bill C-27, which aimed to create Canada’s first comprehensive AI privacy regime, has faced repeated delays and remains under debate. Soft-law instruments like the Treasury Board’s Directive on Automated Decision-Making guide federal agencies but lack binding authority, leaving a regulatory gap for private organizations.

Provincial laws, particularly Quebec’s Law 25, offer a more robust model for ethical ADM. Organizations must notify individuals when decisions rely exclusively on automated processing and provide rights to access, correct, or delete personal information.⁴³ Unlike the GDPR, which applies only to decisions affecting legal status or rights,⁴⁴ Law 25 applies broadly: making human oversight, transparency, and accountability central to compliance. These frameworks demonstrate how privacy and ethics can be integrated into AI systems, emphasizing that ADM must support efficiency without undermining individual rights or public trust.

AuroraAI and human-centric public services in Finland 

Finland has approached AI in government differently from many countries, centering public policy on citizens’ well-being rather than purely on efficiency or industrial competitiveness.⁴⁵ Its AI strategy emphasizes ethical adoption with human-centricity as paramount, and this strategy has earned support through strong collaboration between academia, the private sector, and the Finnish government.⁴⁶ Initiatives like MyData give individuals control over personal information, while free digital skills courses have reached hundreds of thousands of citizens.⁴⁷ These foundational efforts create fertile ground for experimentation with AI in public services, maintaining the crucial aspect of a people-centered perspective to ADM.⁴⁸

AuroraAI, a program run by the Ministry of Finance, exemplifies this human-centered approach.⁴⁵ It is behaviorally designed as a platform where AI organizes public services around individual needs and life events, rather than static bureaucratic processes. The pilot, conducted over five months, used data to match citizens with services relevant to situations such as moving for study or work. By integrating iterative citizen feedback, the program aims to shift public management from output-focused to responsive, personalized service delivery.

The AuroraAI pilot highlighted practical insights on real-world implementation. For example, a “moving to a place of study” life-event pilot surveyed students in Tampere and Turku, revealing that factors like public transport reliability and environmental quality affect well-being. Data from the surveys informed the development of a digital bot that could provide relevant guidance and resources. Uptake was limited, however, showing that even well-designed AI services require proactive marketing and user experience literacy within municipalities to effectively reach citizens.

Scaling AuroraAI presented both promise and challenges. The €100M rollout planned for 2019–2023 had to navigate the complexity of organizational and cultural change within government. Risks included program opacity, data privacy management, and the potential misuse of sensitive information. Lessons from MyData served as a best-practice reference for purpose-limited personal data collection. By 2023, major phases of the rollout had concluded, though debates about its clarity and long-term direction remained.⁴⁹

Similar initiatives in Canada and Poland illustrate different approaches. Canada focuses on immigration efficiency,  Poland on employment service optimization,  and Finland’s model underscores the potential of AI to enhance citizen-centered governance where the interplay between ethics and ADM is prioritized.^{50, 51} Together, these examples highlight that while contexts differ, the global trajectory of ADM experiments points toward a common challenge of balancing innovation with public trust and ethical responsibility.

AI algorithms at work: How to use AI to help overcome historical biases 

Some workplaces are using AI for interviewing, while others are using it to increase fair workplace practices. In this piece, Ariel LaFayette and Turney McKee give an overview of public perception of AI algorithms, historical uses of AI, and important commentary on how AI can combat stereotyping through automated decision-making.

The potential and pitfalls of AI in healthcare

In its ubiquity across industries, AI’s impact is seen all over healthcare from faster waiting room times to better medical diagnoses. In this piece, Sophie Cleff discusses the potential benefits of AI and healthcare, like supporting healthcare workers and better accessibility for patients, as well as its risks, like cognitive bias and data concerns. 

1. Mökander J, Morley J, Taddeo M, Floridi L. Ethics-Based Auditing of Automated Decision-Making Systems: Nature, Scope, and Limitations. Sci Eng Ethics. 2021 Jul 6;27(4):44. doi: 10.1007/s11948-021-00319-4. PMID: 34231029; PMCID: PMC8260507.
2. Aggarwal, N., Eidenmüller, H., Enriques, L., Payne, J., & Van Zwieten, K. (Eds.). (2019). Autonomous systems and the law. CH Beck.
3. Grote, T., & Berens, P. (2020). On the ethics of algorithmic decision-making in healthcare. Journal of medical ethics, 46(3), 205-211.
4. Tutun, S., Johnson, M. E., Ahmed, A., Albizri, A., Irgil, S., Yesilkaya, I., ... & Harfouche, A. (2023). An AI-based decision support system for predicting mental health disorders. Information Systems Frontiers, 25(3), 1261-1276.
5. Cecez-Kecmanovic, D. (2025). Ethics in the world of automated algorithmic decision-making – A Posthumanist perspective. Information and Organization, 35(3), 100587. https://doi.org/10.1016/j.infoandorg.2025.100587
6. Bellamy RKE, Mojsilovic A, Nagar S, Natesan Ramamurthy K, Richards J, Saha D, Sattigeri P, et al. AI fairness 360: An extensible toolkit for detecting and mitigating algorithmic bias. IBM Journal of Research and Development. 2019 doi: 10.1147/JRD.2019.2942287.
7. Gebru, T., Morgenstern, J., Vecchione, B., Vaughan, J. W., Wallach, H., Iii, H. D., & Crawford, K. (2021). Datasheets for datasets. Communications of the ACM, 64(12), 86-92.
8. Koene, A., Clifton, C., Hatada, Y., Webb, H., & Richardson, R. (2019). A governance framework for algorithmic accountability and transparency. Brussels: European Parliamentary Research Service
9. Kim, P. 2017. Auditing algorithms for discrimination. University of Pennsylvania Law Review, 166, 189–203.
10. Ethics guidelines for trustworthy AI. (2019, April 8). European Commission. https://digital-strategy.ec.europa.eu/en/library/ethics-guidelines-trustworthy-ai
11. Pazzanese, C. (2020, October 26). Ethical concerns Mount as AI takes bigger decision-making role. Harvard Gazette. https://news.harvard.edu/gazette/story/2020/10/ethical-concerns-mount-as-ai-takes-bigger-decision-making-role/
12. Shehzadi, T. (2025). Responsible AI: Ethical Considerations in Automated Decision Systems. 10.13140/RG.2.2.17904.98567. 
13. Sanderson, C., Lu, Q., Douglas, D., Xu, X., Zhu, L., & Whittle, J. (2022, December). Towards implementing responsible AI. In 2022 IEEE International Conference on Big Data (Big Data) (pp. 5076-5081). IEEE.
14. What is AI ethics? (2024, September 17). IBM. https://www.ibm.com/think/topics/ai-ethics
15. Eliot, L. (2022, September 3). Unpacking The Best Top Ten Quotes About Artificial Intelligence Leveraging Modern-Day AI Ethics Thinking. Forbes. https://www.forbes.com/sites/lanceeliot/2022/09/03/unpacking-the-best-top-ten-quotes-about-artificial-intelligence-leveraging-modern-day-ai-ethics-thinking/
16. Burrell, J. (2016). How the machine ‘thinks’: Understanding opacity in machine learning algorithms. Big Data & Society, 3(1). https://doi.org/10.1177/2053951715622512 (Original work published 2016)
17. Muth, K. T. (2013, December 3). Bad proxy variables and institutional ethnocentric oversimplification. Global Policy Journal. https://www.globalpolicyjournal.com/blog/03/12/2013/bad-proxy-variables-and-institutional-ethnocentric-oversimplification
18. Dolata, M., Feuerriegel, S., & Schwabe, G. (2022). A sociotechnical view of algorithmic fairness. Information Systems Journal, 32(4), 754-818.
19. Martin, F. (2019, April 22). AI is automated decision-making, and it accelerates century-old algorithmic methods. The CSTA Advocate Blog. https://advocate.csteachers.org/2019/04/22/ai-is-automated-decision-making-and-it-accelerates-century-old-algorithmic-methods/
20. Johnson, D. G. (1997). Ethics and Technology: Ethical Issues in an Age of Information and Communication Technology. Prentice Hall.
21. Sweeney, L. (2000). Simple Demographics Often Identify People Uniquely. Carnegie Mellon University Data Privacy Working Paper 3. https://dataprivacylab.org/projects/identifiability/paper1.pdf
22. Latanya Sweeney. (n.d.). Harvard Kennedy School. https://www.hks.harvard.edu/faculty/latanya-sweeney
23. O’Neil, C. (2016). Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy. Crown.
24. Radden, N. (2020, July). Ethical Issues in any Automated Decision-Making Model. Society of Actuaries. https://www.soa.org/globalassets/assets/library/newsletters/actuarial-technology-today/2020/july/att-2020-07-raden.pdf
25. Buolamwini, J., & Gebru, T. (2018). Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification. Proceedings of Machine Learning Research, 81, 1–15.
26. Churchill, K. (2021, May 27). Foundations of trustworthy AI: Governed data and AI, AI ethics and an open diverse ecosystem. IBM. https://www.ibm.com/think/insights/trustworthy-ai-foundations#:~:text=IBM%27s%20governed%20data%20and%20AI,how%20we%20define%20trustworthy%20AI
27. European Commission. (2024). The Artificial Intelligence Act. Official Journal of the European Union.
28. Bridging Behavioral Science and Artificial Intelligence. (2025). BehavioralAI Institute. https://behavioralai-institute.org/
29. Timnit Gebru. (n.d.). DAIR (Distributed AI Research Institute). https://www.dair-institute.org/team/timnit-gebru/
30. Ferguson, A. G. (2017). The rise of big data policing: Surveillance, race, and the future of law enforcement. The rise of big data policing. New York University Press.
31. Van Brakel, R., & Govaerts, L. (2025). Exploring the impact of algorithmic policing on social justice: Developing a framework for rhizomatic harm in the pre-crime society. Theoretical criminology, 29(1), 91-109.
32. Van Brakel, R., & Govaerts, L. (2025). Exploring the impact of algorithmic policing on social justice: Developing a framework for rhizomatic harm in the pre-crime society. Theoretical criminology, 29(1), 91-109.
33. Završnik, A. (2021). Algorithmic justice: Algorithms and big data in criminal justice settings. European Journal of criminology, 18(5), 623-642.
34. Cheong, B. C. (2024). Transparency and accountability in AI systems: safeguarding wellbeing in the age of algorithmic decision-making. Frontiers in Human Dynamics, 6, 1421273.
35. Floridi, L., Lee, M. S. A., & Denev, A. (2020). Innovating with confidence: embedding AI governance and fairness in a financial services risk management framework.
36. Berman, K. (2025, September 2). 3 lessons from ChatGPT's launch (And what your AI product can learn from it). Substack. https://kristenberman.substack.com/p/3-lessons-from-chatgpts-launch-and?utm_source=substack&publication_id=476445&post_id=172587811&utm_medium=email&utm_content=share&utm_campaign=email-share&triggerShare=true&isFreemail=true&r=1ndm0b&triedRedirect=true
37. Kim, S. S., Liao, Q. V., Vorvoreanu, M., Ballard, S., & Vaughan, J. W. (2024, June). " I'm Not Sure, But...": Examining the Impact of Large Language Models' Uncertainty Expression on User Reliance and Trust. In Proceedings of the 2024 ACM conference on fairness, accountability, and transparency (pp. 822-835).
38. Araujo, T., Helberger, N., Kruikemeier, S. et al. In AI we trust? Perceptions about automated decision-making by artificial intelligence. AI & Soc 35, 611–623 (2020). https://doi.org/10.1007/s00146-019-00931-w
39. What is automated individual decision-making and profiling? (2025, April 22). Information Commissioner's Office (ICO). https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/individual-rights/automated-decision-making-and-profiling/what-is-automated-individual-decision-making-and-profiling/#id2
40. Ethics, transparency and accountability framework for automated decision-making. (2021, May 13). GOV.UK. https://www.gov.uk/government/publications/ethics-transparency-and-accountability-framework-for-automated-decision-making
41. van Bekkum, M., & Borgesius, F. Z. (2021). Digital welfare fraud detection and the Dutch SyRI judgment. European Journal of Social Security, 23(4), 323-340. https://doi.org/10.1177/13882627211031257 (Original work published 2021)
42. Borgesius, F. Z., & Van Bekkum, N. (2021, August 23). Digital welfare fraud detection and the Dutch SyRI judgment. IAPP. https://iapp.org/news/a/digital-welfare-fraud-detection-and-the-dutch-syri-judgment/?utm_source=
43. Millán, L. (2025, February 21). Privacy experts grappling with automated AI decision-making. CBA/ABC National. https://nationalmagazine.ca/en-ca/articles/law/in-depth/2025/privacy-experts%C2%A0grappling-with%C2%A0automated-ai-decision-making
44. General Data Protection Regulation. (2024, April 22). General Data Protection Regulation (GDPR). https://gdpr-info.eu/
45. Kuziemski, M., & Misuraca, G. (2020). AI governance in the public sector: Three tales from the frontiers of automated decision-making in democratic settings. Telecommunications Policy, 44(6), 101976. https://doi.org/10.1016/j.telpol.2020.101976
46. Finland Government. (2019c). Digitalisation and data economy priorities under Finland’s EU Council Presidency.
47. MyData. (n.d.). Personal data management model. https://mydata.org
48. Lomborg, S., Kaun, A., & Scott Hansen, S. (2023). Automated decision-making: Toward a people-centred approach. Sociology Compass, 17(8), e13097. https://doi.org/10.1111/soc4.13097
49. What we learned from AuroraAI: The pitfalls of doing ethics around unsettled technologies. (2023, December 11). FCAI. https://fcai.fi/eab-blog/2023/12/11/what-we-learned-from-auroraai-the-pitfalls-of-doing-ethics-around-unsettled-technologies
50. Villeneuve, S. (2019). Rebooting regulation: Exploring the future of AI Policy in Canada. Brookfield Institute for Innovation Entrepreneurship.
51. Niklas, Jedrzej. (2015). Profiling the Unemployed in Poland: Social and Political Implications of Algorithmic Decision Making.