AI Ethics in the Age of Data: Navigating Risks and Building Constraints
In the digital era, where data is often referred to as the new oil, artificial intelligence (AI) has emerged as a transformative force reshaping industries, economies, and societies. From personalized recommendations on social media platforms to predictive analytics in healthcare and finance, AI systems are increasingly embedded in everyday life. However, as these technologies become more sophisticated and pervasive, they also bring forth a host of ethical challenges—particularly concerning data privacy, algorithmic bias, and the concentration of digital power. These concerns are not merely theoretical; they have real-world implications that affect individual rights, social equity, and democratic processes.
A recent scholarly investigation by Yang Danxiu, an associate professor at the School of Media and Communication at Yunnan University of Finance and Economics, offers a comprehensive analysis of the ethical risks associated with AI-driven information value development. Published in China Media Technology, the study delves into the structural origins of data and algorithmic ethics issues, mapping out how biases, privacy violations, and market imbalances emerge within AI ecosystems. The research underscores the urgent need for a robust, multi-layered ethical constraint mechanism to ensure that technological advancement does not come at the cost of human dignity and social justice.
The foundation of modern AI lies in vast datasets and complex algorithms designed to learn from patterns in human behavior. These systems rely on massive amounts of user-generated data—ranging from search histories and location tracking to social interactions and purchasing habits. While such data enables unprecedented levels of personalization and efficiency, it also opens the door to significant ethical dilemmas. As Yang’s study illustrates, the very process of collecting, analyzing, and applying this data introduces vulnerabilities that can compromise individual autonomy and societal fairness.
One of the central concerns raised in the paper is the issue of data ethics, which encompasses a broad spectrum of problems including privacy breaches, unclear data rights, information monopolies, and the widening digital divide. The complexity begins at the source: data is not neutral. It is shaped by the behaviors, preferences, and biases of both users and collectors. For instance, historical crime statistics used to train predictive policing algorithms may reflect systemic racial profiling rather than actual criminal behavior. As Yang points out, FBI data from 2016 showed African Americans accounting for 37% of reported crimes, a figure that fails to account for decades of institutional discrimination and over-policing in minority communities. When such skewed data is fed into machine learning models, the resulting algorithms perpetuate and even amplify existing social inequalities.
Moreover, the boundaries of data ownership and consent remain murky. Users often unknowingly surrender personal information through opaque terms of service agreements, granting platforms extensive access to their digital footprints. In many cases, these permissions are buried in lengthy legal documents or presented as non-negotiable conditions for using essential services. This lack of transparency undermines informed consent—a cornerstone of ethical data practice. The case of WeChat and Douyin, both operated by Tencent, exemplifies this problem. In 2020, Beijing Internet Court ruled that the apps had violated user privacy by collecting and processing personal data without adequate disclosure or control mechanisms. Similarly, in early 2020, unauthorized dissemination of personal details of individuals returning from Wuhan during the initial phase of the pandemic exposed over 4 million people to potential harassment and identity theft, highlighting the fragility of data protection in times of crisis.
Another critical dimension of data ethics is market imbalance. A handful of tech giants—such as Google, Facebook, Alibaba, and Tencent—control the majority of global data resources, creating what scholars call a “data oligopoly.” This concentration of power allows these companies to dominate advertising markets, influence public opinion, and stifle competition. Smaller platforms struggle to access high-quality training data, placing them at a structural disadvantage. The 2018 incident involving Dianping and Xiaohongshu further illustrates this dynamic: Dianping allegedly used web crawlers to scrape user-generated content from Xiaohongshu, republishing travel notes and reviews without permission. Such practices not only violate intellectual property norms but also distort the competitive landscape, discouraging innovation and harming consumer trust.
Beyond data collection and ownership, the way algorithms interpret and act upon data introduces another layer of ethical complexity. Algorithmic ethics refers to the moral implications of automated decision-making systems—how they are designed, what values they encode, and who bears responsibility when things go wrong. According to Yang, algorithmic risks stem from three primary sources: the inherent unpredictability of machine learning systems, the subjective value preferences embedded by developers, and the absence of comprehensive regulatory frameworks.
Machine learning models, particularly those based on deep neural networks, are often described as “black boxes” because their internal logic is difficult to interpret. These systems can generate novel patterns and make decisions that were not explicitly programmed, leading to outcomes that even their creators cannot fully anticipate. A notable example occurred in 2016 when Facebook’s content moderation algorithm mistakenly flagged Nick Ut’s iconic photograph The Terror of War—a Pulitzer Prize-winning image depicting a naked Vietnamese girl fleeing a napalm attack—as inappropriate and removed it from the platform. The backlash that followed revealed the tension between automated enforcement and cultural sensitivity, underscoring the limitations of rule-based filtering in nuanced contexts.
Even more troubling are the instances where algorithms reflect and reinforce human prejudices. Because AI systems learn from historical data, any societal bias present in that data can be replicated and institutionalized. In 2015, Google Photos’ image recognition system infamously labeled two Black individuals as “gorillas,” sparking widespread condemnation. Though the company quickly issued an apology and attempted to fix the flaw, the incident highlighted how seemingly neutral technology can carry deeply embedded racial assumptions. This phenomenon, known as “bias in, bias out,” occurs when training datasets underrepresent certain groups or contain discriminatory patterns, leading to unfair treatment in areas such as hiring, lending, and law enforcement.
Algorithmic discrimination also manifests in economic practices like “big data price discrimination,” where platforms charge different prices to different users based on their browsing history, location, or purchasing power. In late 2020, Taobao Supermarket faced public scrutiny after users discovered that members of its 88 VIP program were being shown higher prices than regular customers for identical products. This form of dynamic pricing, while profitable for businesses, erodes consumer trust and raises questions about fairness and transparency in digital commerce.
Perhaps one of the most insidious effects of algorithmic curation is the creation of “information cocoons” or “filter bubbles.” Personalized recommendation engines, designed to maximize user engagement, tend to prioritize content that aligns with a person’s existing beliefs and interests. Over time, this leads to intellectual isolation, where individuals are rarely exposed to diverse perspectives or challenging ideas. The consequences extend beyond personal enrichment; they threaten the foundations of democratic discourse. During the 2016 U.S. presidential election, Cambridge Analytica leveraged data from 50 million Facebook profiles to deliver micro-targeted political advertisements, exploiting psychological profiles to manipulate voter behavior. The scandal revealed how algorithmic systems could be weaponized to influence public opinion, undermine electoral integrity, and polarize societies.
Given the scale and severity of these ethical challenges, Yang argues that a passive approach to AI governance is no longer tenable. Instead, a proactive, multi-dimensional constraint mechanism must be established—one that integrates legal regulation, industry self-governance, technological innovation, and public awareness. Such a framework should not aim to stifle innovation but to channel it toward socially beneficial ends.
At the legislative level, clear and enforceable data protection laws are essential. The European Union’s General Data Protection Regulation (GDPR), implemented in 2018, serves as a model for comprehensive privacy legislation. By granting individuals rights such as data access, correction, erasure (“right to be forgotten”), and portability, GDPR empowers users to exert greater control over their personal information. It also imposes strict obligations on organizations regarding data transparency, accountability, and breach notification. While China has made progress with the Personal Information Protection Law (PIPL) enacted in 2021, there remains room for refinement in enforcement mechanisms and cross-sectoral coordination.
Complementing legal frameworks, industry self-regulation plays a crucial role in fostering responsible AI development. Professional associations, standard-setting bodies, and corporate ethics boards can establish best practices for data collection, algorithmic auditing, and impact assessment. For example, tech companies should adopt “privacy by design” principles, embedding data protection measures into the architecture of their products from the outset. They should also implement algorithmic transparency reports, disclosing how decisions are made and allowing independent researchers to evaluate potential biases. Internal ethics review committees, similar to institutional review boards in medical research, could assess high-risk AI applications before deployment.
Technological solutions must also evolve to address the root causes of ethical failures. One promising direction is value-sensitive design—a methodology that integrates ethical considerations into the engineering process. Developers should conduct scenario-based risk assessments during the design phase, anticipating how their systems might be misused or produce unintended consequences. Techniques such as adversarial testing, where models are deliberately fed misleading inputs to expose vulnerabilities, can help identify and mitigate biases before deployment. Additionally, explainable AI (XAI) tools that provide interpretable outputs can enhance accountability and enable users to understand and challenge automated decisions.
Public education and digital literacy are equally important. As AI systems become more integrated into daily life, citizens must develop the skills to critically evaluate algorithmic recommendations, recognize manipulation tactics, and advocate for their rights. Schools, media outlets, and civil society organizations should promote awareness of data rights and algorithmic influences. Empowering users with knowledge reduces dependency on opaque systems and fosters a more informed and resilient digital citizenry.
Yang emphasizes that ethical AI is not just a technical challenge but a collective societal endeavor. It requires collaboration across disciplines—computer science, law, philosophy, sociology, and psychology—to build systems that reflect shared human values. The goal is not to eliminate all risks, which is impossible, but to create a balance between innovation and responsibility, efficiency and equity, convenience and autonomy.
Looking ahead, the evolution of AI will likely introduce new ethical frontiers. As systems grow more autonomous, questions about agency, liability, and consciousness will intensify. Should a self-driving car prioritize passenger safety over pedestrian lives in an unavoidable accident? Who is accountable when an AI-generated artwork infringes on copyright? How should societies regulate artificial general intelligence if it ever emerges? These are not hypotheticals for distant futures—they are emerging realities that demand foresight and preparedness.
In conclusion, the rapid advancement of artificial intelligence presents both extraordinary opportunities and profound ethical challenges. As demonstrated by Yang Danxiu’s research, the risks associated with data exploitation and algorithmic decision-making are deeply rooted in technical, economic, and cultural structures. Addressing them requires more than patchwork fixes; it demands a systemic transformation in how we design, deploy, and govern intelligent technologies. By establishing robust legal safeguards, promoting industry accountability, advancing ethical engineering practices, and enhancing public understanding, society can harness the benefits of AI while safeguarding fundamental rights and democratic values. The path forward is not one of technological determinism, but of intentional choice—choosing to build a future where intelligence serves humanity, not the other way around.
Yang Danxiu, School of Media and Communication, Yunnan University of Finance and Economics. Published in China Media Technology. DOI: 10.19483/j.cnki.11-4653/n.2021.07.019