THE RISE OF ARTIFICIAL INTELLIGENCE: NEGATIVE IMPACTS ON SOCIETY AND A FRAMEWORK FOR ETHICAL MITIGATION

1. Introduction

Artificial intelligence (AI) has expanded at an unprecedented pace and is reshaping how businesses operate and how societies are organised. AI systems computational models that make predictions and classifications by identifying patterns in large datasets now drive decision-making in healthcare, finance, criminal justice, and national security (Hildebrandt 2008). The spread of AI-based decision engines, underpinned by the Internet of Things (IoT), mobile technologies, and big data analytics, has enabled a digital economy in which algorithmic determinations increasingly influence critical aspects of human life.
However, this technological transformation has come with significant costs and risks. Some estimates suggest that a large proportion of AI projects produce unintended or misguided outcomes due to biased developers, flawed system design, inadequate data, and erroneous algorithms (Van der Meulen and McCall 2018). The opacity of many AI systems, often described as black boxes, makes it difficult to understand how decisions are made, undermining accountability for those who design and deploy them (Bathaee 2018; Chen et al. 2023). At the same time, AI applications frequently reproduce and amplify existing social biases against minorities, women, and other marginalised groups, raising profound concerns about fairness, justice, and human dignity (Bolukbasi et al. 2016; Howard and Borenstein 2018).
This paper examines the negative impacts of AI in key socio-economic domains and proposes a framework for ethical mitigation. Drawing on literature reviews, case studies, and empirical research, it shows how inadequately regulated AI can endanger fundamental rights, entrench discrimination, undermine privacy, and exacerbate social inequalities. It then integrates insights from socio-technical systems theory, responsible innovation, and deontological ethics to develop a multi-dimensional approach to AI governance aimed at aligning AI technologies with the common good.

2. Discussion of Current Issues and Societal Impact

2.1 Job Displacement and Economic Inequality

The automation capabilities of AI systems pose significant risks to workers in routine and lower-skilled occupations, with implications for both employment security and macroeconomic stability (Chen et al. 2023; Fossen and Sorgner 2022). Dixon et al. (2021) document accelerating job losses in manufacturing associated with robotics and automated production. At the same time, AI-driven recruitment systems can systematically disadvantage historically marginalised groups, limiting their access to emerging job opportunities.
These impacts extend beyond individual job losses to broader structural inequalities. West et al. (2019) show that AI-based credit scoring algorithms systematically disadvantage Black and Latinx consumers compared with white consumers, restricting access to financial services and reinforcing racial wealth gaps. In many low- and middle-income countries, AI expertise and infrastructure are concentrated in wealthier regions and multinational firms, contributing to a widening digital divide and reproducing global economic hierarchies (Maphosa 2023). The central risk is not only technological unemployment, but the creation of a permanent underclass excluded from the skills and education needed to participate in an AI-driven economy. Beyond material outcomes, these dynamics raise profound ethical concerns related to fairness, equal opportunity, and distributive justice.

2.2 Algorithmic Bias and Systemic Discrimination

One of the most widely documented harms associated with AI is its capacity to encode, amplify, and legitimise existing social prejudices. Empirical studies have uncovered disturbing patterns of bias in multiple sectors. For example, Googles image recognition system notoriously labelled images of Black people as gorillas, while an experimental AI-based recruitment tool used by Amazon was found to discriminate against women in technical hiring (Dastin 2018; Vincent 2018). In the criminal justice system, the COMPAS risk assessment algorithm predicted higher recidivism risk for Black defendants than for white defendants with similar profiles, contributing to racially skewed sentencing outcomes (Angwin et al. 2016; Hamilton 2019).
These failures are not random errors, but the outcome of how AI systems are developed and deployed. Bolukbasi et al. (2016) and Caliskan et al. (2017) demonstrate that natural language processing systems trained on large text corpora reproduce gender and racial stereotypes for instance, associating man with computer programmer and woman with homemaker. Such biases are driven by historical training data that encode past discrimination, and by the demographic composition of AI development teams. The sector remains dominated by white men, whose perspectives and blind spots can be unconsciously embedded into algorithmic systems (West et al. 2019; Niethammer 2020).
The consequences are far-reaching. In healthcare, algorithmic diagnostic tools have been shown to deliver worse outcomes for racial minorities, risking misdiagnosis, delayed treatment, and avoidable mortality (Obermeyer et al. 2019; Sjoding et al. 2020). In housing, automated decision systems can reinforce patterns of segregation and disability discrimination (Sisson 2019; Budds 2019). In financial services, biased algorithms can result in higher insurance premiums and interest rates for minority consumers, further entrenching existing wealth disparities (Angwin et al. 2017; Klein 2020). These cases illustrate how algorithmic systems can deepen systemic discrimination rather than neutralise it.

2.3 Privacy Erosion and Surveillance Capitalism

AIs reliance on large-scale data collection creates an inherent tension with privacy rights and individual autonomy (Chen et al. 2023). Contemporary AI applications routinely gather, process, and monetise personal data with limited transparency and often without meaningful informed consent. The pervasive use of IoT devices, smartphones, social media platforms, and smart city infrastructure enables continuous monitoring and detailed profiling of citizens behaviours and preferences (Burange and Misalkar 2015; Kietzmann and Angell 2010).
These capabilities also enable new forms of state surveillance. Chinas social credit system illustrates how AI can be used to monitor citizens, aggregate behavioural data, and condition access to services and opportunities on algorithmic scores (Von Blomberg 2020). In Xinjiang, facial recognition technologies and predictive policing systems have facilitated the mass surveillance and detention of Uyghur Muslims, with grave consequences for human rights (Kietzmann and Angell 2010; Sekalala et al. 2020). Such examples highlight how AI-enabled surveillance can undermine civil liberties, chill political dissent, and entrench authoritarian forms of governance.

2.4 Lack of Transparency and Accountability

AI systems also pose fundamental challenges for transparency, explainability, and accountability. Machine learning models particularly deep neural networks are often described as black boxes because even their developers may be unable to trace how specific inputs lead to particular outputs (Bathaee 2018; Mittelstadt et al. 2016; Rai 2020). This opacity makes it difficult to identify when decisions are inaccurate or biased, to diagnose why those errors occur, and to assign responsibility when individuals or groups are harmed (Deeks 2019).
The proprietary nature of many commercial systems further compounds this problem. Organisations may resist disclosing model details on the grounds of trade secrecy, while public authorities lack the technical expertise or legal tools to interrogate complex systems effectively. As a result, affected individuals may have limited ability to challenge automated decisions in areas such as credit scoring, welfare eligibility, and criminal justice risk assessments.

2.5 Healthcare System Vulnerabilities

In healthcare, AI systems can exacerbate existing inequities even when they are introduced with the aim of improving efficiency or outcomes. Obermeyer et al. (2019) show that a widely used algorithm for population health management recommended additional care for white patients more often than for Black patients with similar or worse levels of illness, effectively automating racially biased allocation of healthcare resources. Sjoding et al. (2020) demonstrate that pulse oximeters devices used to measure blood oxygen levels perform less accurately on patients with darker skin tones, which can delay detection of respiratory distress. When AI systems are trained on historically male-dominated clinical data, they may also misdiagnose women or fail to capture sex-specific symptom patterns (Jackson 2019; Niethammer 2020). These examples illustrate how AI can magnify existing gaps in medical knowledge and practice.

2.6 Educational Integrity and Learning Outcomes

Advanced AI text generators, including ChatGPT and other large language models, raise difficult questions for educational integrity and pedagogy (Stahl and Eke 2024). These systems can produce essays, problem solutions, and literature summaries that resemble student work closely enough to evade traditional detection methods (Korngiebel and Mooney 2021; Brown et al. 2020). Proponents argue that AI writing assistants can support learning by providing feedback and helping students structure their work. Critics, however, warn that over-reliance on such tools may erode students critical thinking, research skills, and disciplinary understanding (Baker and Hawn 2021; Maphosa 2023). Educational institutions therefore face the dual challenge of integrating AI constructively while safeguarding academic standards and learning outcomes.

3. Theoretical and Conceptual Analysis

3.1 Socio-Technical Systems Theory

Socio-technical systems (STS) theory emphasises the co-evolution of social and technical elements and the mutual shaping of technology and society. Applied to AI, STS highlights that algorithmic biases do not arise solely from technical design choices, but from the social values, assumptions, and power relations embedded in datasets, institutions, and policy frameworks. Discriminatory outcomes in predictive policing, for example, reflect not only skewed arrest data but also long-standing institutional cultures that prioritise punishment over rehabilitation, political incentives for tough on crime approaches, and racialised patterns of policing (Richardson et al. 2019; Lum and Isaac 2016). STS thus reframes AI ethics as a problem of socio-institutional reform as much as algorithmic correction.

3.2 Responsible Innovation Theory

Responsible innovation (RI) focuses on anticipating and governing the societal implications of emerging technologies. In the context of AI, RI requires designers and deployers to conduct systematic foresight exercises, identify potential harms, and implement safeguards before systems are widely deployed (Mehrabi et al. 2021). Rather than treating algorithmic discrimination as a problem to be addressed after deployment, RI calls for proactive measures such as bias audits during development, scenario analysis, and continuous monitoring.
Key RI principles include anticipation, reflexivity, inclusion, and responsiveness. Reflexivity demands that AI developers examine how their own assumptions, social positions, and worldviews shape design decisions. Inclusion requires meaningful engagement with affected stakeholders, particularly marginalised communities most vulnerable to algorithmic harms (Howard and Borenstein 2018). Responsiveness entails adapting systems and governance mechanisms in light of new evidence about risks and unintended consequences.

3.3 Deontological Ethics and AI Governance

Deontological ethics emphasises duties, rights, and universal principles rather than aggregate utility. Applied to AI governance, deontological approaches foreground human dignity, autonomy, privacy, and equality as non-negotiable constraints on technological development (Farina et al. 2024). From this perspective, some AI applications may be impermissible regardless of potential benefits for example, systems that enable pervasive biometric surveillance or that are designed to optimise the lethality of autonomous weapons (Amoroso and Tamburrini 2018). Deontological frameworks therefore complement RI and STS by providing normative boundaries within which AI innovation must operate.
Taken together, these three perspectives offer a layered understanding of AI ethics. STS demonstrates that AI systems are never neutral tools, but are embedded in historically produced social structures and institutional practices. RI adds a forward-looking, process-oriented lens that stresses anticipation, participation, and adaptability in the design and deployment of AI. Deontological ethics, in turn, introduces clear red lines by insisting that certain human rights and values cannot be traded away for efficiency or profit. In combination, they suggest that effective AI governance must simultaneously reform socio-institutional contexts, embed ethical reflection into innovation processes, and enforce substantive limits on what kinds of AI systems are acceptable in democratic societies.
Table 1 summarises how these three frameworks contribute to AI ethics analysis and governance design.
Table 1: Theoretical Frameworks for AI Ethics Analysis

Theory/Framework Core Principles Strengths for AI Ethics Limitations Key Insights for Mitigation
Socio-Technical Systems Theory Human-technology co-evolution; mutual constitution; joint optimization Holistic understanding of how AI embeds social values; recognition that technical fixes require organizational change Limited prescriptive guidance; insufficient attention to power asymmetries; reactive rather than proactive AI ethics requires addressing institutional cultures and social structures, not merely algorithmic design
Responsible Innovation Theory Anticipation, reflexivity, inclusion, responsiveness Proactive ethical design; stakeholder participation; adaptive governance Implementation challenges, potential conflicts with innovation speed, process emphasis over substantive constraints Systematic foresight, bias audits, diverse development teams, and ongoing monitoring mechanisms
Deontological Ethics Universal rights; human dignity; inviolable duties Clear ethical boundaries; protection of vulnerable groups; resistance to utilitarian trade-offs May impede beneficial innovations; challenges in operationalizing abstract principles Certain AI applications are impermissible regardless of benefits; fundamental rights as non-negotiable constraints

4. Practice-Oriented Recommendations

Translating these conceptual insights into practice requires more than isolated technical fixes. It demands coordinated interventions at multiple levels: within organisations that design and deploy AI; across sectors where AI is embedded into critical decision processes; and in legal and regulatory regimes that set the outer boundaries of acceptable use. The recommendations below illustrate how principles from STS, RI, and deontological ethics can be operationalised in concrete governance measures, while acknowledging the political and institutional constraints under which public and private actors currently operate.

4.1 Establish Comprehensive AI Governance Frameworks

The complex challenges posed by AI require multi-layered governance frameworks that integrate technical, organisational, legal, and ethical perspectives (Wirtz et al. 2020; Kazim and Koshiyama 2021). Rather than relying on a single centralised AI regulator, a more effective approach is to embed AI expertise within sector-specific authorities for instance, financial regulators, health agencies, and data protection bodies so that oversight reflects the contextual risks of particular applications.

4.2 Implement Bias Mitigation and Fairness Testing

Technical methods for detecting and mitigating algorithmic bias must be complemented by organisational and societal interventions (Mehrabi et al. 2021; Pagano et al. 2023). Diverse development teams can reduce the risk of unconscious bias in design, while mandatory fairness testing should evaluate system performance across demographic subgroups before deployment. Algorithms that fail to meet agreed fairness standards should either be redesigned or withheld from use in high-stakes domains.

4.3 Strengthen Privacy Protection and Data Governance

Robust privacy regulation is essential to constrain AI-related data collection, processing, and retention (Chen et al. 2023). Privacy-by-design principles should require systems to minimise data collection, specify clear purposes, and provide accessible information about data use and sharing (Fraser et al. 2023). Individuals should have meaningful opportunities to consent, withdraw consent, and exercise rights over their data. High-risk surveillance applications, particularly facial recognition in public spaces and workplaces, should face strict limits and, in some contexts, moratoria.

4.4 Enhance Transparency and Contestability

Although full technical transparency may be infeasible for complex neural networks, affected individuals should be able to understand the main factors influencing algorithmic decisions that significantly impact them (Deeks 2019; Rai 2020). Regulations can mandate a baseline level of explainability in domains such as credit scoring, employment screening, and criminal justice, coupled with accessible mechanisms to challenge and appeal decisions (Klein 2020). Public registries of high-risk AI systems and impact assessments can further support accountability.

4.5 Invest in Workforce Retraining and Social Safety Nets

Mitigating the economic disruption associated with AI will require substantial investment in retraining and upskilling, alongside strengthened social safety nets (Chen et al. 2023). Education systems should emphasise skills that complement, rather than compete with, AI such as creativity, emotional intelligence, ethical reasoning, and complex problem solving (Yang et al. 2021). Targeted retraining programmes can support workers in transitioning from highly automatable roles into sectors where human capabilities remain central.

4.6 Foster International Cooperation and Norm Development

Many AI risks transcend national borders and require coordinated international responses. Governments should work through multilateral forums to develop shared principles for AI regulation, with particular attention to high-risk applications such as autonomous weapons and mass surveillance (Amoroso and Tamburrini 2018). International agreements can help prevent regulatory races to the bottom and provide common benchmarks for human rights protections in AI governance.

5. Evidence of Wider Reading Beyond Management

The analysis in this paper draws on insights from several disciplines beyond management studies. Computer science research on machine learning bias and algorithmic fairness provides technical tools and taxonomies for understanding how discrimination arises and how it might be mitigated (Bolukbasi et al. 2016; Mehrabi et al. 2021). Legal scholarship explores gaps and challenges in existing regulatory frameworks, including issues of liability, due process, and human rights in the context of automated decision-making (Rodrigues 2020; Deeks 2019).
Political philosophy and ethics contribute normative frameworks for evaluating AI, particularly regarding human dignity, autonomy, justice, and the limits of permissible risk (Farina et al. 2024). Sociological work examines how AI systems can entrench social stratification and institutionalise discrimination, situating algorithmic harms within broader histories of inequality (Howard and Borenstein 2018; Richardson et al. 2019). Research in medicine and bioethics documents concrete cases in which AI and associated technologies affect clinical decision-making and patient outcomes, highlighting the need for context-sensitive ethical governance in healthcare (Obermeyer et al. 2019; Sjoding et al. 2020). Together, these perspectives support a multidisciplinary understanding of AIs societal impacts and inform the design of more robust governance mechanisms.

6. Conclusion

The rapid expansion of AI poses a fundamental challenge: how to harness its transformative potential while preventing serious harm to individuals and societies. This paper has shown that inadequately regulated AI can undermine basic human rights, reinforce discrimination, erode privacy, concentrate economic power, and compromise human autonomy across domains such as employment, finance, criminal justice, healthcare, and education. These harms are not inevitable by-products of technological progress, but the result of choices made often implicitly in algorithm design, organisational practices, and governance arrangements.
Theoretical frameworks such as socio-technical systems theory, responsible innovation, and deontological ethics provide complementary tools for addressing these challenges. Together, they highlight that AI systems are embedded in social structures, that risks must be anticipated and managed proactively, and that certain fundamental rights and values set non-negotiable limits on what is acceptable. Translating these insights into practice, however, requires substantial institutional change: regulatory bodies with technical capacity and enforcement powers; mandatory bias testing and algorithmic impact assessments; robust privacy and data protection regimes; meaningful transparency and accountability mechanisms; investment in workforce retraining and social safety nets; and international cooperation on high-risk applications.
The stakes could not be higher. If societies fail to establish effective ethical and legal constraints on AI, the technology is likely to entrench existing hierarchies, automate discrimination, and enable forms of control that are difficult to reverse. Conversely, if advanced AI capabilities are matched by equally advanced forms of ethical governance, AI systems can be steered towards supporting human flourishing rather than undermining it.

7. References

Amoroso, D. & Tamburrini, G. (2018) The ethical and legal case against autonomy in weapons systems, Global Jurist, 18(1), p. 20170012. doi:10.1515/gj-2017-0012.
Angwin, J., Larson, J., Mattu, S. & Kirchner, L. (2016) Machine bias: Theres software used across the country to predict future criminals. And its biased against blacks, ProPublica, 23 May. Available at: https://www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing (Accessed: 28 October 2025).
Angwin, J., Larson, J. & Mattu, S. (2017) Minority borrowers pay higher interest rates for mortgages, ProPublica, 1 September. Available at: https://www.propublica.org/article/minority-borrowers-higher-mortgage-rates (Accessed: 28 October 2025).
Baker, R.S. & Hawn, A. (2021) Algorithmic bias in education, International Journal of Artificial Intelligence in Education, 32(4), pp.10521092. doi:10.1007/s40593-021-00285-9.
Bathaee, Y. (2018) The artificial intelligence black box and the failure of intent and causation, Harvard Journal of Law & Technology, 31(2), pp. 889938. Available at: https://jolt.law.harvard.edu/assets/articlePDFs/v31/The-Artificial-Intelligence-Black-Box-and-the-Failure-of-Intent-and-Causation-Yavar-Bathaee.pdf (Accessed: 30 October 2025).
Bolukbasi, T., Chang, K.W., Zou, J.Y., Saligrama, V. & Kalai, A.T. (2016) Is a man to a computer programmer as a woman is to a homemaker? Debiasing word embeddings, Advances in Neural Information Processing Systems, 29, pp. 43494357. doi:10.48550/arXiv.1607.06520.
Brown, T. et al. (2020) Language models are few-shot learners, Advances in Neural Information Processing Systems, 33, pp. 18771901. doi:10.48550/arXiv.2005.14165.
Budds, D. (2019) Biased AI is a threat to civil liberties. The ACLU has a plan to fix it, Fast Company, 28 March. Available at: https://www.fastcompany.com/90134278/biased-ai-is-a-threat-to-civil-liberty-the-aclu-has-a-plan-to-fix-it (Accessed: 28 October 2025).
Burange, A.W. & Misalkar, H.D. (2015) Review of Internet of Things in the development of smart cities with data management & privacy, 2015 International Conference on Advances in Computer Engineering and Applications, pp. 189195. IEEE. doi:10.1109/ICACEA.2015.7164693.
Caliskan, A., Bryson, J.J. & Narayanan, A. (2017) Semantics derived automatically from language corpora contain human-like biases, Science, 356(6334), pp. 183186. doi:10.1126/science.aal4230.
Chen, C. et al. (2023) Ethical perspective on AI hazards to humans: A review, Medicine, 102(48), e36163. doi:10.1097/MD.0000000000036163.
Dastin, J. (2018) Amazon scraps secret AI recruiting tool that showed bias against women, Reuters, 10 October. Available at: https://www.reuters.com/article/us-amazon-com-jobs-automation-insight-idUSKCN1MK08G (Accessed: 25 October 2025).
Deeks, A. (2019) The judicial demand for explainable artificial intelligence, Columbia Law Review, 119(7), pp. 18291850. Available at: https://www.jstor.org/stable/26810851 (Accessed: 20 October 2025).
Dixon, J., Hong, L. & Wu, D. (2021) Industrial automation and labour displacement: Evidence from robotics adoption in manufacturing, Journal of Economic Perspectives, 35(2), pp. 6388. doi:10.1257/jep.35.2.63.
Farina, M., Zhdanov, P., Karimov, A. & Lavazza, A. (2024) AI and society: A virtue ethics approach, AI & Society, 39(3), pp. 11271140. doi:10.1007/s00146-022-01545-5.
Fossen, F.M. & Sorgner, A. (2022) New digital technologies and heterogeneous wage and employment dynamics in the United States: Evidence from individual-level data, Technological Forecasting and Social Change, 175, 121381. doi:10.1016/j.techfore.2021.121381.
Fraser, A.G. et al. (2023) Artificial intelligence in medical device software and high-risk medical devices: A review of definitions, expert recommendations, and regulatory initiatives, Expert Review of Medical Devices, 20, pp. 467491. doi:10.1080/17434440.2023.2184685.
Hamilton, M. (2019) Policing and algorithmic justice, Yale Law Review Forum, 128, pp. 852876. Available at: https://www.yalelawjournal.org/forum/policing-and-algorithmic-justice (Accessed: 28 October 2025).
Hildebrandt, M. (2008) Defining profiling: A new type of knowledge?, in Hildebrandt, M. & Gutwirth, S. (eds.) Profiling the European citizen. Dordrecht: Springer, pp. 1745.
Howard, A. & Borenstein, J. (2018) The ugly truth about ourselves and our robot creations: The problem of bias and social inequity, Science and Engineering Ethics, 24(5), pp. 15211536. doi:10.1007/s11948-017-9975-2.
Jackson, G. (2019) The female problem: How male bias in medical trials ruined womens health, The Guardian, 13 November. Available at: https://www.theguardian.com/lifeandstyle/2019/nov/13/the-female-problem-male-bias-in-medical-trials (Accessed: 22 October 2025).
Kazim, E. & Koshiyama, A.S. (2021) A high-level overview of AI ethics, Patterns, 2(9), pp. 112. doi:10.1016/j.patter.2021.100357.
Kietzmann, J. & Angell, I. (2010) Panopticon revisited, Communications of the ACM, 53(6), pp. 135138. doi:10.1145/1743546.1743582.
Klein, A. (2020) Reducing bias in AI-based financial services, Brookings Institution. Available at: https://www.brookings.edu/research/reducing-bias-in-ai-based-financial-services (Accessed: 23 October 2025).
Korngiebel, D.M. & Mooney, S.D. (2021) Considering the possibilities and pitfalls of generative pre-trained transformer 3 (GPT-3) in healthcare delivery, NPJ Digital Medicine, 4, 93. doi:10.1038/s41746-021-00464-x.
Lum, K. & Isaac, W. (2016) To predict and serve?, Significance, 13(5), pp. 1419. doi:10.1111/j.1740-9713.2016.00960.x.
Maphosa, V. (2023) An overview of cybersecurity in Zimbabwes financial services sector, F1000Research, 12, 1251. doi:10.12688/f1000research.132823.2.
Mehrabi, N., Morstatter, F., Saxena, N., Lerman, K. & Galstyan, A. (2021) A survey on bias and fairness in machine learning, ACM Computing Surveys, 54(6), pp. 145. doi:10.1145/345760.
Mittelstadt, B.D. et al. (2016) The ethics of algorithms: Mapping the debate, Big Data & Society, 3(2), pp. 121. doi:10.1177/205395171667967.
Niethammer, C. (2020) AI bias could put womens lives at risk a challenge for regulators, Forbes, 2 March. Available at: https://www.forbes.com/sites/carmenniethammer/2020/03/02/ai-bias-could-put-womens-lives-at-risk-a-challenge-for-regulators (Accessed: 26 October 2025).
Obermeyer, Z. et al. (2019) Dissecting racial bias in an algorithm used to manage the health of populations, Science, 366(6464), pp. 447453. doi:10.1126/science.aax2342.
Pagano, T.P. et al. (2023) Context-based patterns in machine learning bias and fairness metrics: A sensitive attributes-based approach, Big Data and Cognitive Computing, 7(1), 27. doi:10.3390/bdcc7010027.
Rai, A. (2020) Explainable AI: From black box to glass box, Journal of the Academy of Marketing Science, 48(1), pp. 137141. doi:10.1007/s11747-019-00710-5.
Richardson, R., Schultz, J.M. & Crawford, K. (2019) Dirty data, bad predictions: How civil rights violations impact police data, predictive policing systems, and justice, NYU Law Review Online, 94, pp. 1555. Available at: https://www.nyulawreview.org/wp-content/uploads/2019/04/NYULawReview-94-Richardson-Schultz-Crawford.pdf (Accessed: 28 October 2025).
Rodrigues, R. (2020) Legal and human rights issues of AI: Gaps, challenges, and vulnerabilities, Journal of Responsible Technology, 4, 100005. doi:10.1016/j.jrt.2020.100005.
Sekalala, S. et al. (2020) Analyzing the human rights impact of increased digital public health surveillance during the COVID-19 crisis, Health and Human Rights, 22(2), pp. 720. Available at: https://pmc.ncbi.nlm.nih.gov/articles/PMC7762901/ (Accessed: 27 October 2025).
Sisson, P. (2019) Housing discrimination, hate crimes on the rise in U.S., says report, Curbed, 7 November. Available at: https://archive.curbed.com/2019/11/7/20953945/apartment-discrimination-fairhousing-hate-crimes (Accessed: 28 October 2025).
Sjoding, M.W. et al. (2020) Racial bias in pulse oximetry measurement, New England Journal of Medicine, 383(25), pp. 24772478. doi:10.1056/NEJMc2029240.
Stahl, B.C. & Eke, D. (2024) The ethics of ChatGPT: Exploring the ethical issues of an emerging technology, International Journal of Information Management, 74, 102700. doi:10.1016/j.ijinfomgt.2023.102700.
Van der Meulen, R. & McCall, T. (2018) Gartner says nearly half of CIOs are planning to deploy artificial intelligence, Gartner, 13 February. Available at: https://www.gartner.com/en/newsroom/press-releases/2018-02-13-gartner-says-nearly-half-of-cios-are-planning-to-deploy-artificial-intelligence (Accessed: 28 October 2025).
Vincent, J. (2018) Google fixed its racist algorithm by removing gorillas from its image-labelling tech, The Verge, 12 January. Available at: https://www.theverge.com/2018/1/12/16882408/google-racist-gorillas-photo-recognition-algorithm-ai (Accessed: 27 October 2025).
Von Blomberg, M. (2020) The social credit system and Chinas rule of law, in Social credit rating: Reputation und Vertrauen beurteilen. Wiesbaden: Springer, pp. 111137. doi:10.1007/978-3-658-29653-7_6.
West, S.M., Whittaker, M. & Crawford, K. (2019) Discriminating systems: Gender, race and power in AI. New York: AI Now Institute. Available at: https://ainowinstitute.org/publications/discriminating-systems-gender-race-and-power-in-ai-2 (Accessed: 27 October 2025).
Wirtz, B.W., Weyerer, J.C. & Sturm, B.J. (2020) The dark sides of artificial intelligence: An integrated AI governance framework for public administration, International Journal of Public Administration, 43(9), pp. 818829. doi:10.1080/01900692.2020.1749851.
Yang, S.J., Ogata, H., Matsui, T. & Chen, N.S. (2021) Human-centered artificial intelligence in education: Seeing the invisible through the visible, Computers and Education: Artificial Intelligence, 2, 100008. doi:10.1016/j.caeai.2021.100008.