The Ethical Implications of Artificial Intelligence
Artificial intelligence refers to technology designed to think and behave like humans. This can be manifested in various ways, including understanding human speech, playing chess or video games, interpreting complex data, and learning.
AI includes machine learning, a form of AI that trains systems to learn patterns in data. When developed and used in an ethical manner, the application of AI has the potential to significantly improve quality of life, add value to organizations, and create innovation.
Conversely, the misuse of AI has profound and far-reaching consequences for individuals and society. It can result in discrimination, violate privacy, pose a danger to human life or safety, and may pose a security risk.
AI systems are created, directed, and used by people and, as such, should be used in ways consistent with our ethics, values, and societal laws. Discussions that provoke thinking about the deep influence of AI are ongoing, including the inherent capacity of AI to make decisions about objectives and how they will be pursued.
These decisions will have consequences in value-based domains such as finance, health, and criminal justice. In these settings, AI will help decide who gets a loan, promotion, or, for instance, who is entitled to health care or a good education.
Ethics is concerned with the principles of right and wrong that govern the behavior and decision-making of an individual, group, or system. Unlike traditional coded software, which follows a strict set of instructions, AI can learn and alter its behavior based on the data it is designed to interpret.
However, AI itself cannot make moral decisions because it has no moral reasoning capacity; it is amoral. For example, if AI is trained on data with implicit bias, such as human resources practices, the AI could make biased decisions autonomously. That is why there is an ethical obligation for developers to be mindful of the implications of their work.
People involved in a decision about the development, implementation, and use of AI are responsible if that decision brings either harm or good. Those who have a significant effect on a decision about what technological artifact we should build or impose on society have an ethical responsibility.
This includes technical engineers, business stakeholders, managers, governments, human-computer interaction designers, and regulatory agencies. They bear a particular level of responsibility to ensure their AI systems are used in ethically informed choices.
The Code of Ethics and Code of Professional Conduct details the responsibility of technologists in ensuring their work does no harm, that it is legally compliant, and that it is of overall benefit to society.
Defining Artificial Intelligence
Artificial Intelligence (AI) is a vast concept that has fascinated researchers for over 50 years. The 'intelligence' that makes AI unique takes many forms, from the narrow intelligence similar to AI, which focuses on typical tasks such as autonomous driving and intelligent trading, to the general AI that represents higher cognitive qualities like those observed in humans.
The capability of AI-enabled computers, electronic gadgets, or robots to mimic cognitive functions like logical reasoning, problem-solving, image recognition, voice, and language interpretation is exceeding in terms of speed, scope, and complexity.
AI operates on high-end technical algorithms, complex arithmetic, and data logic with high processing capabilities, machine learning, natural language processing, intelligent data mining, searching and learning, and generally focuses more on behavior than reasoning. AI technologies and applications are rapidly evolving and will soon be seamlessly integrated into various aspects of our personal and professional lives.
Traditional AI systems will be based on a problem-solving method using heuristic search and knowledge representation to solve complex challenges. AI concepts and applications have gradually led to the adoption of some popular methods such as supervised learning, unsupervised learning, semi-supervised learning, and reinforcement learning for various AI applications.
They not only learn from knowledge and datasets, but they also change in decision-making to respond to new threats and opportunities. The capabilities of AI led to the foundation of pure AI, strong AI, and cognitive science to educate and empower AI systems with cognitive reasoning and problem facilitation.
AI today is a set of algorithms that can use different technologies to perform various tasks. An AI-powered computer can, for example, perform specialized actions based on logical reasoning, if-then decisions, behaviors, image recognition, or natural language processing.
There are typically two types of AI: strong AI or super AI and weak AI or sub AI. Today’s AI applications are really narrow AI, which assists in carrying out and managing specific tasks, but in the future, AI will be converted into general and even super AI.
Importance of Ethical Considerations
As artificial intelligence (AI) based systems become deeply integrated into various aspects of human life, ethical considerations become paramount in building human-friendly AI systems and in directing their use towards beneficial outcomes.
AI affects various aspects of human life, and ethical considerations are needed to ensure the systems are used to influence society and daily life in a manner that causes the least societal harm. Building ethical AI systems is vital for consumer trust, which is crucial for the success of AI in the economy.
The ethical deployment of these systems will thus be a part of the system’s commercial value proposition. Industry, policymakers, and academia alike agree on the importance of human-centered AI, the protection of fundamental rights, such as personal data, privacy, dignity, and freedom, and the provision of transparency, responsibility, and accountability.
AI and machine learning tools are used in decision systems, and they can contribute to reinforcing existing societal biases and patterns of discrimination by generating and operationalizing patterns of inequality.
Irrespective of the place of operation of the AI, the potential to do harm is a characteristic of AI. Once again, a proactive approach must be taken not only to mitigate potential harm but also to create potential for social benefit.
Overall, our approach is to bundle ethical issues with the development of technology, which is undertaken to showcase how all actors contributing to the development and deployment of the technology need to be aware of issues that can bear repercussions on a societal level.
Field experts are increasingly defining the understanding and raising awareness of AI ethics outside of traditional places of discussion. There is a necessity for multidisciplinary theories to ensure diverse backgrounds and diverse thought trends are taken into account in understanding the societal effect of AI.
AI ethics is an umbrella structure that showcases the involvement of all concerned parties in the ethical bearing of AI. The aim is to bring forward awareness of ethical dilemmas that need to be addressed in AI systems. In the following, we name those primarily engaged in the development and deployment processes concerned stakeholders.
Historical Context of AI Ethics
Artificial intelligence (AI) has been in development in some form since the concept was first proposed in the 1940s. In 1956, the first AI program was capable of winning a game of chess, and the first mobile robots could think and sense by the end of the decade.
Between 1968 and 1977, the United States government founded AI research centers, while the 1980s and 1990s saw even more advances in computing and research, leading to new machine learning algorithms and new methods of solving difficult computer problems more efficiently.
Today, AI can do anything from simulating human reasoning to making decisions at automated car washes, in combination with incarcerating suspects using facial recognition software. This doesn't capture the scale or implications of AI, just a small slice of its capacities and how it is being used.
While AI has undergone iterations and improvements, AI ethics has grown parallel to it. The relevance, incidents, research, responsibility, and policy phases are the stages of AI ethical determination that have been identified, a roadmap informed in part by AI's history.
In the era of relevance, AI was a concept born in academic philosophy. During this time, three main problems inspired debate: theory of mind (can AI think and feel like a human?), intentionality (is AI capable of doing things and thinking about them as if it had beliefs and desires?), and consequences (consider the consequences of treating AI in (1) and (2)).
The lucky phrase raised was "mindless operations following rules." AI's capacity to outmatch the finest living chess competitor at the time was a significant outcome of the logic of mindless regulations, among other results.
Milestones in AI Development
1945: A hypothetical machine called a Memex that could store, organize, and retrieve information such as books, files, and communications is described. The Memex has often been described as the conceptual forerunner of the World Wide Web.
1950: A test for determining whether a computer can truly think is proposed. A computer could be said to have artificial intelligence if it could trick a human interrogator into believing that the written responses to its questions were generated by another human.
1951: The world’s first commercially available general-purpose programmable computer is developed. The machine was nearly 5 meters long, weighed more than a ton, and cost about $415,000 in today’s dollars. It could store just 3,000 decimal digits in its memory.
1956: A conference is organized, credited as the birth of modern artificial intelligence (AI). The workshop resulted in the first academic work related to AI research in the English language.
1967: During a public demonstration, a computer-controlled wheelchair arrives at a specified location in a hallway and navigates through a narrow corridor. It was an early example of machine learning and a harbinger of today’s self-driving cars.
1973: A press release suggests that there may be “no postmasters, mail sorters, or special delivery letter carriers” by 1985 because such jobs could be performed by machines using technology similar to the zip code readers then being tested. This was among the first large-scale public discussions of the potential for widespread technological unemployment, long before AI became a household term.
1979-1980: An annual competition for computer programs that can play games is sponsored. The first games are backgammon, followed by chess the next year. The winners produced a combination of rule-learning algorithms and large-scale database searching that represented the beginning of commercial AI.
An AI program wins a game of chess against a world champion. It was the first time a game of this complexity ended in victory for a computer. A question-answering system that can learn and process natural language is developed. In 2015, it beats two reigning champions.
In 2019, the creators of a system very similar to the chess-playing system start a business, with the goal of creating a similar chess-playing system except that it could learn to play chess at the level of a grandmaster without using historical data. After just one day of practice, the AI is reportedly able to beat the leading chess engine.
In 2018, another version of its software that learns to play a game through a process of trial and error, without human instruction, is released.
Emergence of Ethical Concerns
As the AI field continued to advance, societies began to see AI-powered systems being used to make important decisions. With this increasing dependence on these systems, experts and non-experts alike began to question the fairness, accountability, and transparency of such "black-box" technologies.
This resonated with cases from prior years, in which a system that used ML to predict flu outbreaks from analyzing search results was overpredicting flu outbreaks. In subsequent years, these debates have been echoed with others regarding fairness and AI decision-making, including a resume screening tool developed by a major company that was later challenged with legal cases.
Academia, industry, and governments took note of these incidents, beginning to engage in specialized conferences focused on addressing these generated ethical dilemmas. The development of workshops to help familiarize researchers with the ethical issues raised by AI are both born of this response.
Additionally, there was increased public concern over the ethical implications of AI in 2018, the same year that significant data protection regulations were passed and implemented. Proponents of AI argue that these advances in technology can help to empower individual employees in negotiations with employers.
As seen, the same technology that can empower individuals also raises critical ethical challenges. This suggests that for AI systems to responsibly integrate into society, it will be crucial to develop strategies for accurately assessing these unique ethical features of new technologies or tools.
Key Ethical Principles in AI
The fundamental principles for guiding the ethical implications of artificial intelligence (AI) have been discussed at length, leading to the adoption of responsible AI frameworks by international organizations.
As researchers and developers move forward with the creation of new AI systems, it is important to take a good look at the principles that make up these frameworks. Transparency and explainability refer to the responsibility of AI developers to clarify how AI systems make decisions and why, in a way that is understandable to potential users.
Furthermore, it relates to how potential users may be informed about the reliability, strengths, and limitations of the AI decision-making process. Fairness and bias touch on the responsibility of AI developers to address and limit bias in a way that avoids discrimination, ensuring equitable AI decision-making towards people and communities.
Lastly, privacy and data protection relate to AI developers' responsibility to ensure that the data inputs to AI systems are managed in a secure and responsible manner, free from unauthorized access or misuse.
These principles are intertwined, and neglecting any one has the potential to threaten others. Indeed, a lack of trust in the transparency and ethical handling of sensitive data for algorithm training reduces potential adopters' willingness to engage with it and lowers public confidence in the fairness of the AI system outcomes.
Ethical principles of responsible AI can, therefore, pave the way for meaningful evaluation or assessment of AI developments that can reconcile technical innovation with public acceptability and societal values.
It also opens up the potential for the operational implementation of these principles in the design and deployment of AI applications with greater public value. AI demands serious ethical reflection in order to contribute enduring practical value to society as a whole.
The world faces the risk that AI developments could face a plethora of socio-politically motivated regulatory barriers, owing to the fear of researchers and developers of the technology being seen as indifferent about the pain their technology inflicts upon the world.
While the growing traction of public bodies, including governments, intergovernmental and industry standards organizations towards responsible AI is a welcome sign, it raises further questions about the concrete methods for integrating ethical principles into the practice of AI.
It is hoped that this argument brings attention to these ethical principles and re-emphasizes those geared around social consequences and advances reflections and potential ways to grapple with ethical AI in practice and regulation.
Transparency and Explainability
Introduction User trust is essential and in the interest of everyone involved when using AI. Transparency in AI refers to the extent users are capable of understanding how certain AI systems work and how decisions are made by them, a crucial aspect for assessing how accountable and answerable these systems can be.
The more impactful the potential outcomes of decisions made by the system are, or in other words, the higher the stakes, the more necessary these standards become. The AI systems make decisions from data, so it follows that the ability of the user to understand the decision must stem from an understandability of the outputs these systems generate. Such a property has been recently defined as explainability in AI.
Various problems stem from the opaqueness of algorithms. The first is one of societal trust. If people do not understand why a specific piece of music was recommendable or not, they will not trust the system making the recommendation.
Trust in systems, particularly to the appropriate degree in high-stakes decisions, is considered central to societal uptake of AI. The second risk is illicit use by specific operators. If an operator of an AI system has designed it so that no one really understands how it works, then the potential for hiding underlying non-legitimate operations increases, something that may be used to bypass legally guaranteed non-discriminatory algorithms.
Such a tendency is of particular interest in the context of technology transfer, as it raises some issues of normativity. The idea that systems pose risks is not new, but it is a point that we suggest provides some initial guidelines in defining appropriate transparency and interpretability standards for AI systems, as it suggests the need for having systems that work in similar ways as those that are allowable.
In particular, we believe that in social science applications and, in particular, technology transfer applications, it is necessary that the data and data treatment processes fed into suggesting a next step correspond to actually used possible next steps in current employment practices. But this is difficult to judge without some process of explanation in place.
Fairness and Bias
Privacy and Data Protection
In many setups, AI applications require data. When it comes to applications like recommender systems or advanced analytics, this requirement could mean user profiling. In turn, user profiling could potentially lead to privacy risks, such as restricted freedom, discrimination, identity theft, and data misuse. These privacy risks are one element of the ethical concerns surrounding data collection for AI.
Related ethical issues might include the consent for data collection, data ownership, and potential data misuse or lock-in effects. Despite these ethical concerns and ongoing public debate, an analysis of a representative sample of the Alexa Top 1,000 domains revealed that roughly 81.9% of those websites use client-side tracking.
In Europe, the General Data Protection Regulation ensures that the processing of personal data needs to be transparent, fair, lawful, and accountable. These are further steps in addressing ethical issues regarding data collection and AI-related privacy problems.
Furthermore, the European focus in these documents is primarily on the impacts of AI on privacy. In practice, major social networking platforms, search engines, and businesses have been using data for years with the goal of developing better AI.
However, their privacy practices have been continuously part of a public debate and discussions about ethical AI. Consequently, privacy and data protection are ethical considerations not only if AI systems rely on personal data, but also if AI could contribute to a potential misuse of general data. AI systems, for instance, might be used for surveillance purposes or to support fake news.
Ultimately, loss of trust is one argument in debating the balance between using data for the goal of a better AI that might let companies innovate and secure the economy and considering personal rights.
This is because a loss of trust might have negative social consequences. In a position statement, a working group criticizes the balancing act and pleads for respecting human rights. They open the paper by stating, 'Our overriding concern is to carefully craft policies for responsible AI that indeed serve the public good, and not just any entity claiming to embody that good.' To them, being smart is different from being ethical. Without privacy, 'not only human freedom and dignity, but also social schemes are troubled.' Ensuring privacy is needed for a 'climate of trust'.
Ethical Dilemmas in AI Decision-Making
Automation has the potential to greatly improve decision-making in a wide range of areas. However, as AI systems become a component of, or in some cases the decision-making agent, there is an increasing need to address ethical issues.
The increasing ability for a decision to be fully automated raises questions about the potential ethical dilemmas created by automated decisions, particularly as these decisions start to encroach on human life. These issues are particularly contentious where the stakes surrounding a decision could be high, such as in criminal justice, politics, healthcare, finance, and customer service, or where the decision will affect a large group of people.
Various studies have discussed the potential of AI, and in particular machine learning, to produce undesirable behavior, which could range in severity from poor decisions to actions that are damaging to others.
Accountability and transparency are two further issues in the design of ethical AI systems. Often the level of autonomy given to a decision-making system, AI or otherwise, dictates its level of control and is likely to inform the development of ethical guidelines.
In situations such as finance, where the effects of decisions are cascaded throughout the system, legislators are increasingly interested in making AI explainable and accountable. This is required in these cases as the system cannot demonstrate ethical agency in any other way.
Fairness is a well-documented issue in the development of AI, and it is particularly relevant when the output of a system is making a value-laden decision or represents an act of discrimination. It is especially concerning that low representation groups, who traditionally experience higher rates of discrimination, are less likely to trust AI, so any decision to limit or withdraw human input may create particularly unfair decisions. It has been shown that given the wrong data, AI may learn about minorities from historical data in a derogatory way, and it is deemed unethical and inefficient to deploy AI vis-à-vis humans without the necessary adjustments.
When AI reproduces, or even exaggerates, these biases, it can result in decisions that will perpetuate and even exacerbate the bias. The introduction of autonomous AI severely exacerbates the situation as existing ethical concerns about human involvement are magnified when humans are removed from the decision-making process.
There is a potential for humans to lose trust in black box systems taking automatic and independent decisions in any of these sectors. The tension between data collection and societal expectations about privacy also needs to be addressed in relation to the deployment of ethical AI decision-making systems. Nonandrocentric ethics about AI in male-dominated careers are still heavily engaging with the ethics of
AI from a Western perspective, with limited critical analysis of what this knowledge creates, how it should be used, and to what ends. In all scenarios, robust ethical frameworks need to be agreed upon prior to the deployment of AI, which accounts for the limitations and mapping between technological developments and the responsibility of AI in decision-making.
The filtering of these dilemmas is crucially about societal disclosure of the limitations of AI as a morally obligated decision engine, autonomous decision-making process authority, and even as an assistant.
By ethically assessing the role of AI in decision-making, we are prompted to consider the exact function of this technology and closely align the decision to the ethics of the organizational body. New technologies that promote ethical decision-making will be less likely to negatively impact at-risk groups.
Further, understanding the moral dimension of self-improving AI could improve the fairness and efficiency of AI decision-making. Ultimately, AI and its role in society necessitate ongoing ethical reflection on whether automation is the correct course of action in these decision fields.
Autonomous Vehicles
Intelligent machines enter, leaving the scene for trolley problems. Industry and academia have been engaged in a lively debate around the ethics of life and death decision-making by AVs, akin to the famous philosophical thought experiment called the trolley problem.
Following certain ethical principles, many accounts of ethical machines potentially include protecting the interests of others or minimizing harm. In tradocratic parlance, the 'trolley problem' of AVs can be understood as asking if an AV should stick to its path and sacrifice its passengers for the greater good.
AV behavior in such accident avoidance scenarios is influencing public opinion on AV safety. Different public perceptions from varying societies surround the moral acceptability of lethal AVs in terms of pedestrian and passenger safety.
The response to the moral issues of autonomous driving depends on society's values and the acceptable level of risk due to self-driving. Philosophy lends essential insights into how we may or should behave as the developers of autonomous systems.
It is guided by reflections on what it means to act ethically, morally, and with integrity in the realm of autonomous technology. Essentially, it encourages those in the field of automotive AI to think hard about how we make these systems – before we even consider deploying these technologies on city streets.
There is a rich field of ethical reflection on these issues that researchers and roboticists have much to gain. It is important to discuss the moral and ethical implications of the use of AI and autonomy, as well as the development of such technology.
Healthcare Applications
The use of AI in healthcare can help to rapidly and accurately provide a diagnosis and care pathway, potentially improving health outcomes for patients. It can also be used to eliminate menial tasks that burden staff, and it allows the collection and analysis of extensive medical data.
However, there are concerns that such large quantities of sensitive patient data are being used without either the explicit or implicit consent of patients. Ethical AI involves transparency and appropriate consent.
Increasing developments in medicine are leading to an increase in personalized medicine, and diagnosis and treatment decisions are likely to become increasingly complex. When AI applications are used in diagnostics, they will make decisions about what is present in the scan obtained about a human without their informed consent—either directly in relation to consent to the information being gained or indirectly where consent is not sought about the information being stored or used.
Mechanisms will be needed to ensure transparency so that a user is able to audit what data was used in achieving an outcome and how the utility of that data affects the decision.
The use of healthcare AI systems could inadvertently support decision-making processes that have primarily been based on stereotypes and discrimination of existing healthcare professionals. If employed, many AI systems will use the data of populations to reach their recommendations and decisions about further diagnoses and treatments.
These decisions, however, may reflect current discriminatory practices and thereby support the further proliferation of inequality in healthcare, which will lead to a reinforcement of discriminatory and unequal health outcomes.
Legislators, companies, and related stakeholders in the creation, use, and regulation of AI need to work within an ethical framework where patient welfare is central to decisions over the creation and use of AI for diagnostic procedures.
The AI community also ought to include consideration of diagnosis as a human-centered task and take human considerations and the moral judgments that AI-enabled diagnostics demand into account. Patients may currently expect that a doctor would oversee their diagnosis. They ought to expect the same if that process involves using AI.
Regulatory Frameworks and Guidelines
National and International Regulations
National regulations on artificial intelligence: A number of countries and international actors have produced AI governance frameworks, largely based on the principle of responsible AI. The most prominent ones are the United States, the European Union, the United Kingdom, Canada, Singapore, and South Korea.
The United States recently published its own strategy, which focuses on government affairs and industry competitiveness. The global technological arms race between the United States and China brings international attention, with national security and economic futures at stake. Therefore, international cooperation can be limited since nations adopt different ethical priorities based on their own historical and cultural backgrounds.
International regulations: International actors are also designating ethical AI standards. The most notable actors are the European Union and the Organization for Economic Cooperation and Development, which have proposed AI guidelines.
The European Union guidelines focus on autonomy, mentorship, intergenerational responsibility, and peace and security. The OECD’s AI Policy Observatory has also established its own AI Principles and created the dataset on AI Public Policy Instruments.
The main objective is to collect data that shows AI principles, standards, regulations, guidelines, and governance frameworks, adopted and currently under negotiation or consultation. Furthermore, the United Nations has published eight principles for artificial intelligence and has proposed a Model AI Governance Act.
As might be expected, concerted international collaboration among these agreements has not been successful yet. As a result, they fluctuate between soft and hard laws. Administratively, before they are agreed upon between nations, they must undergo a lengthy parliamentary process.
Furthermore, enforcing their decisions against enforceable or hard laws is difficult or impossible since they lack global recognition. As such, coordination between several organizations is challenging in practice.
For private parties and stakeholders, having too many determinants is difficult to follow and enforce. It is, therefore, important to identify generally accepted ethical theories and to collaborate multilaterally in addressing overwhelming challenges. This includes drafting binding international law alongside various international actors and experts.
Regulatory alignment refers to the concept of developing a collection of procedures from one or more sets of methods. These frameworks, in the case of AI ethics, should establish a universal norm that accommodates different cultural, social, and ethical approaches across countries.
To that end, the implementation of a task force or independent coalition is crucial. The plurilateral group would adhere to the fundamental principles of the political participation model of constitution-making involving experts, business, and the public.
In conclusion, the role of governments should also only be partially restricted to pursuing nationality in regulating AI ethics because their task is also to sanction certain fundamental universal norms. To bring nations together with common values, it is important to seek consensus for the general good.
Those principles should be codified in a globally acknowledged agreement which would filter through into national laws. Nonetheless, more should be adopted on the implementation and enforcement of mechanisms, including a true global governance body, before multinationalism.
Established experts should do this through the Identification Treaty. To summarize, AI governance principles will offer a steer in the direction of AI adoption concerning ethical scarcity. There are already competing groups, each with their set of values. Nonetheless, it seems that this appreciation emphasizes Western theories at the price of individual nations. Therefore, it is crucial to build common and dynamic AI ethics among nations.
Industry Standards and Best Practices
In addition to their internal guidelines, organizations are also looking to industry standards and best practices to help steer and justify their actions. Engineering and other professional societies are based upon a membership of individuals rather than organizations in many cases, but a number of these organizations have taken an important role in shaping ethical standards and efforts for their members.
The AI Standards, for instance, is general guidance that provides five overarching objectives for the ethical standard to strive towards. Organizations have also developed best practices with input from a variety of members, including academia, big tech, and regulatory bodies, which build upon the guides and work done to date.
The Policy on Artificial Intelligence is one of the largest collaborations, spanning industry, the technical community, and governmental regulatory bodies. Section one is a general introduction to the topic, but sections two through four cover a range of ethical guidelines in the development and deployment of AI.
The document on Algorithmic Bias Considerations offers guidance on how to discover, identify, and comprehensively manage multiple types of unfairness resulting from AI decision-making.
Based on input and discussions from workgroups that span industry, academia, and civil society, this document offers a consensus on best practices and gives implementers a way to operationalize fairness in AI systems. Algorithmic fairness is a large topic, and as such, this guidance was written for those seeking high-level considerations of algorithmic fairness.
Other sections cover topics such as data asset management and data quality; data privacy, protection, and sharing; maximizing user happiness and/or quality using explainable and transparent techniques; procedures and evidential standards; and organizational risk management.
While such standards set best practices, they are often intended to guide all manner of social, corporate, and regulatory legislative actors, and are necessarily high-level for this reason. They will need continued adaptation to changing technological and political contexts, as well as to the roles and necessities of the various actors guided by the principles in the ecosystem.
For best practices to work as a guardian against the negative consequences of harmful AI, these need to be partially able to act as an implementer’s best practices guide. To effectively guard against the negative consequences of AI, any of the paths suggested must be developed in conjunction with integrating them as a part of a wider regulatory and stakeholder framework; otherwise, free riders may undercut the efforts of any lone organization’s good behavior.
References:
- Mahajan, S., 2023. Artificial Intelligence and its Impacts on the Society. Contemporary Social Sciences. jndmeerut.org
- Cheng, L., Varshney, K. R., & Liu, H., 2021. Socially responsible ai algorithms: Issues, purposes, and challenges. Journal of Artificial Intelligence Research. jair.org
- Aithal, P.S., 2023. Super-Intelligent Machines-analysis of developmental challenges and predicted negative consequences. International Journal of Applied Engineering and Management Letters (IJAEML), 7(3), pp.109-141. ssrn.com
- Tokayev, K. J., 2023. Ethical implications of large language models a multidimensional exploration of societal, economic, and technical concerns. International Journal of Social Analytics. norislab.com
- Ferrer, X., Van Nuenen, T., Such, J.M., Coté, M. and Criado, N., 2021. Bias and discrimination in AI: a cross-disciplinary perspective. IEEE Technology and Society Magazine, 40(2), pp.72-80. [PDF]
- Keles, S., 2023. Navigating in the moral landscape: analysing bias and discrimination in AI through philosophical inquiry. AI and Ethics. [HTML]
- Ferrara, E., 2023. Fairness and bias in artificial intelligence: A brief survey of sources, impacts, and mitigation strategies. Sci. mdpi.com
- Haider, I., Khan, M.T., Khan, M.S., Saeed, R. and Saleem, S., 2023. ACADEMIC WORKSHOP AND IT’S IMPACT ON LEARNING. Journal of Khyber College of Dentistry, 13(3), pp.7-12. journalofkcd.com
- Wheatley, A. & Hervieux, S., 2022. Separating artificial intelligence from science fiction: Creating an academic library workshop series on AI literacy. mcgill.ca
- Olney, A. M., Chounta, I. A., Liu, Z., Santos, O. C., & Bittencourt, I. I., 2024. Artificial Intelligence in Education: Posters and Late Breaking Results, Workshops and Tutorials, Industry and Innovation Tracks, Practitioners, Doctoral …. [HTML]
- Schiff, D., Borenstein, J., Biddle, J. and Laas, K., 2021. AI ethics in the public, private, and NGO sectors: A review of a global document collection. IEEE Transactions on Technology and Society, 2(1), pp.31-42. techrxiv.org
- de Laat, P. B., 2021. Companies committed to responsible AI: From principles towards implementation and regulation?. Philosophy & technology. springer.com
- Ahmad, A., 2024. Ethical implications of artificial intelligence in accounting: A framework for responsible ai adoption in multinational corporations in Jordan. International Journal of Data and Network Science. growingscience.com
- Mohamed, Y. H. A., 2023. Comprehending and mitigating feature bias in machine learning models for ethical AI. International Journal of Social Analytics. norislab.com
- Agarwal, R., Bjarnadottir, M., Rhue, L., Dugas, M., Crowley, K., Clark, J. and Gao, G., 2023. Addressing algorithmic bias and the perpetuation of health inequities: An AI bias aware framework. Health Policy and Technology, 12(1), p.100702. [HTML]
- Venkatasubbu, S. and Krishnamoorthy, G., 2022. Ethical Considerations in AI Addressing Bias and Fairness in Machine Learning Models. Journal of Knowledge Learning and Science Technology ISSN: 2959-6386 (online), 1(1), pp.130-138. jklst.org
- Jackson, M. C., 2021. Artificial intelligence & algorithmic bias: the issues with technology reflecting history & humans. J. Bus. & Tech. L.. umaryland.edu
- Hacker, P., Mittelstadt, B., Borgesius, F.Z. and Wachter, S., 2024. Generative discrimination: What happens when generative ai exhibits bias, and what can be done about it. arXiv preprint arXiv:2407.10329. [PDF]
- Drage, E. & Frabetti, F., . The performativity of AI-powered event detection: How AI creates a racialized protest and why looking for bias is not a solution. Science. sagepub.com
- Ahmed, F., Fattani, M. T., Ali, S. R., & Enam, R. N., 2022. Strengthening the bridge between academic and the industry through the academia-industry collaboration plan design model. Frontiers in Psychology. frontiersin.org
- Zhuang, T. & Liu, B., 2022. Sustaining higher education quality by building an educational innovation ecosystem in China—policies, implementations and effects. Sustainability. mdpi.com
- Rossoni, A.L., de Vasconcellos, E.P.G. and de Castilho Rossoni, R.L., 2024. Barriers and facilitators of university-industry collaboration for research, development and innovation: a systematic review. Management Review Quarterly, 74(3), pp.1841-1877. springer.com
- Rane, N.L., Kaya, O. and Rane, J., 2024. Human-centric artificial intelligence in industry 5.0: Enhancing human interaction and collaborative applications. Artificial Intelligence, Machine Learning, and Deep Learning for Sustainable Industry, 5, pp.2-95. researchgate.net