Roger Clarke's Web-Site
© Xamax Consultancy Pty Ltd, 1995-2021
|Identity Matters||Other Topics||Waltzing Matilda||What's New|
Revision of 15 April 2019, with editorial corrections of 26 August
(Addition of JSAI, Sony, AustGovt, EC 2019, and Update of IBM, MS)
Prepared in support of Guidelines for the Responsible Business Use of AI
This supersedes the version of 10 February 2019
Roger Clarke **
© Xamax Consultancy Pty Ltd, 2018-19
Available under an AEShareNet licence or a Creative Commons licence.
This document is at http://www.rogerclarke.com/EC/GAIP.html
During the current round of industry enthusiasm for Artificial Intelligence (AI), ambitious claims by technology providers have stimulated widespread public concern. IT suppliers, user organisations in business and government, and associations representing them, are naturally concerned about the prospect of regulation constraining their activities. During the period 2016-19, there has accordingly been a concerted effort by a wide variety of organisations to calm the public's nerves. This has included the publication of 'principles' and 'guidelines' for the implementation of AI. In addition, a number of sets of principles have been published by advocates for the public interest.
Most collections have been put together by, or depend heavily on, organisations that have a vested interest in developing, investing in and/or applying AI. Most collections have involved little or no effective engagement with advocates for the interests of the public. Moreover, the documents impose no actual obligations on any organisation to do or not do anything, and are not capable of being enforced. Any influence they have will derive from the hovering threat of deep public disquiet.
On the other hand, many of these documents have been developed by well-resourced organisations that have access to researchers, developers and implementors of various AI technologies. In extracting the documents' information content, considerable care is needed, in order to appreciate sub-texts, to consider why statements are framed as they are, to understand the effects of qualifying words, and to identify aspects that are entirely missing. Provided that such care is brought to the activity, there is a great deal of value to be extracted from these documents.
This document includes citations to and excerpts from 22 such documents. This document is complemented by a collection of 8 further documents that present principles arising from more general ethical analysis of IT's impacts. Together, the collection of 30 sets of principles provides a basis for a consolidated super-set of 50 Principles for Responsible AI, published in Clarke (2019). Each of the 22 documents in this set is given a score showing how many of the 50 Principles are at least modestly reflected in the document.
1) Research Goal: The goal of AI research should be to create not undirected intelligence, but beneficial intelligence.
2) Research Funding: Investments in AI should be accompanied by funding for research on ensuring its beneficial use, including thorny questions in computer science, economics, law, ethics, and social studies, such as:
3) Science-Policy Link: There should be constructive and healthy exchange between AI researchers and policy-makers.
4) Research Culture: A culture of cooperation, trust, and transparency should be fostered among researchers and developers of AI.
5) Race Avoidance: Teams developing AI systems should actively cooperate to avoid corner-cutting on safety standards.
6) Safety: AI systems should be safe and secure throughout their operational lifetime, and verifiably so where applicable and feasible.
7) Failure Transparency: If an AI system causes harm, it should be possible to ascertain why.
8) Judicial Transparency: Any involvement by an autonomous system in judicial decision-making should provide a satisfactory explanation auditable by a competent human authority.
9) Responsibility: Designers and builders of advanced AI systems are stakeholders in the moral implications of their use, misuse, and actions, with a responsibility and opportunity to shape those implications.
10) Value Alignment: Highly autonomous AI systems should be designed so that their goals and behaviors can be assured to align with human values throughout their operation.
11) Human Values: AI systems should be designed and operated so as to be compatible with ideals of human dignity, rights, freedoms, and cultural diversity.
12) Personal Privacy: People should have the right to access, manage and control the data they generate, given AI systems' power to analyze and utilize that data.
13) Liberty and Privacy: The application of AI to personal data must not unreasonably curtail people's real or perceived liberty.
14) Shared Benefit: AI technologies should benefit and empower as many people as possible.
15) Shared Prosperity: The economic prosperity created by AI should be shared broadly, to benefit all of humanity.
16) Human Control: Humans should choose how and whether to delegate decisions to AI systems, to accomplish human-chosen objectives.
17) Non-subversion: The power conferred by control of highly advanced AI systems should respect and improve, rather than subvert, the social and civic processes on which the health of society depends.
18) AI Arms Race: An arms race in lethal autonomous weapons should be avoided.
19) Capability Caution: There being no consensus, we should avoid strong assumptions regarding upper limits on future AI capabilities.
20) Importance: Advanced AI could represent a profound change in the history of life on Earth, and should be planned for and managed with commensurate care and resources.
21) Risks: Risks posed by AI systems, especially catastrophic or existential risks, must be subject to planning and mitigation efforts commensurate with their expected impact.
22) Recursive Self-Improvement: AI systems designed to recursively self-improve or self-replicate in a manner that could lead to rapidly increasing quality or quantity must be subject to strict safety and control measures.
23) Common Good: Superintelligence should only be developed in the service of widely shared ethical ideals, and for the benefit of all humanity rather than one state or organization.
1 Contribution to humanity
Members of the JSAI will contribute to the peace, safety, welfare, and public interest of humanity. They will protect basic human rights and will respect cultural diversity. As specialists, members of the JSAI need to eliminate the threat to human safety whilst designing, developing, and using AI.
2 Abidance of laws and regulations
Members of the JSAI must respect laws and regulations relating to research and development, intellectual property, as well as any other relevant contractual agreements. Members of the JSAI must not bring harm to others through violation of information or properties belonging to others. Members of the JSAI must not use AI with the intention of harming others, be it directly or indirectly.
3 Respect for the privacy of others
Members of the JSAI will respect the privacy of others with regards to their research and development of AI. Members of the JSAI have the duty to treat personal information appropriately and in accordance with relevant laws and regulations.
Members of the JSAI will always be fair. Members of the JSAI will acknowledge that the use of AI may bring about additional inequality and discrimination in society which did not exist before, and will not be biased when developing AI. Members of the JSAI will, to the best of their ability, ensure that AI is developed as a resource that can be used by humanity in a fair and equal manner.
As specialists, members of the JSAI shall recognize the need for AI to be safe and acknowledge their responsibility in keeping AI under control. In the development and use of AI, members of the JSAI will always pay attention to safety, controllability, and required confidentiality while ensuring that users of AI are provided appropriate and sufficient information.
6 Act with integrity
Members of the JSAI are to acknowledge the significant impact which AI can have on society. They will therefore act with integrity and in a way that can be trusted by society. As specialists, members of the JSAI will not assert false or unclear claims and are obliged to explain the technical limitations or problems in AI systems truthfully and in a scientifically sound manner.
7 Accountability and Social Responsibility
Members of the JSAI must verify the performance and resulting impact of AI technologies they have researched and developed. In the event that potential danger is identified, a warning must be effectively communicated to all of society. Members of the JSAI will understand that their research and development can be used against their knowledge for the purposes of harming others, and will put in efforts to prevent such misuse. If misuse of AI is discovered and reported, there shall be no loss suffered by those who discover and report the misuse.
8 Communication with society and self-development
Members of the JSAI must aim to improve and enhance society's understanding of AI. Members of the JSAI understand that there are diverse views of AI within society, and will earnestly learn from them. They will strengthen their understanding of society and maintain consistent and effective communication with them, with the aim of contributing to the overall peace and happiness of mankind. As highly-specialized professionals, members of the JSAI will always strive for self-improvement and will also support others in pursuing the same goal.
9 Abidance of ethics guidelines by AI
AI must abide by the policies described above in the same manner as the members of the JSAI in order to become a member or a quasi-member of society.
(1) Governance frameworks, including standards and regulatory bodies, should be established to oversee processes assuring that the use of A/IS does not infringe upon human rights, freedoms, dignity, and privacy, and of traceability to contribute to the building of public trust in A/IS.
(2) A way to translate existing and forthcoming legal obligations into informed policy and technical considerations is needed. Such a method should allow for differing cultural norms as well as legal and regulatory frameworks.
(3) For the foreseeable future, A/IS should not be granted rights and privileges equal to human rights, A/IS should always be subordinate to human judgment and control.
A/IS should prioritize human well-being as an outcome in all system designs, using the best available, and widely accepted, well-being metrics as their reference point. [The discussion appears to be primarily concerned with economic wellbeing]
(1) Legislatures/courts should clarify issues of responsibility, culpability, liability, and accountability for A/IS where possible during development and deployment (so that manufacturers and users understand their rights and obligations).
(2) Designers and developers of A/IS should remain aware of, and take into account when relevant, the diversity of existing cultural norms among the groups of users of these A/IS.
(3) Multi-stakeholder ecosystems should be developed to help create norms (which can mature to best practices and laws) where they do not exist ... (including representatives of civil society, law enforcement, insurers, manufacturers, engineers, lawyers, etc.).
(4) Systems for registration and record-keeping should be created so that it is always possible to find out who is legally responsible for a particular A/IS. Manufacturers/operators/ owners of A/IS should register key, high-level parameters, including Training data/training environment (if applicable), Sensors/real world data sources, Algorithms, Process graphs, Model features (at various levels), User interfaces, Actuators/outputs, Optimization goal/loss function/reward function
Develop new standards that describe measurable, testable levels of transparency, so that systems can be objectively assessed and levels of compliance determined. For designers, such standards will provide a guide for self-assessing transparency during development and suggest mechanisms for improving transparency.
Minimize the risks of misuse of A/IS by raising public awareness, providing ethics education, and educating government, lawmakers and enforcement agencies [but with no mention of obligations, sanctions and enforcement]
A WEF document claims that these "core principles" derive from a report commissioned by the House of Lords AI Select Committee, which is based on evidence from over 200 industry experts - most of whom presumably has at least a degree of self-interest in the outcome.
The first principle argues that AI should be developed for the common good and benefit of humanity.
The report's authors argue the United Kingdom must actively shape the development and utilisation of AI, and call for "a shared ethical AI framework" that provides clarity against how this technology can best be used to benefit individuals and society.
They also say the prejudices of the past must not be unwittingly built into automated systems, and urge that such systems "be carefully designed from the beginning, with input from as diverse a group of people as possible".
The second principle demands that AI operates within parameters of intelligibility and fairness, and calls for companies and organisations to improve the intelligibility of their AI systems.
"Without this, regulators may need to step in and prohibit the use of opaque technology in significant and sensitive areas of life and society", the report warns.
Third, the report says artificial intelligence should not be used to diminish the data rights or privacy of individuals, families or communities.
It says the ways in which data is gathered and accessed need to be reconsidered. This, the report says, is designed to ensure companies have fair and reasonable access to data, while citizens and consumers can also protect their privacy.
"Large companies which have control over vast quantities of data must be prevented from becoming overly powerful within this landscape. We call on the government ... to review proactively the use and potential monopolisation of data by big technology companies operating in the UK".
The fourth principle stipulates all people should have the right to be educated as well as be enabled to flourish mentally, emotionally and economically alongside artificial intelligence.
For children, this means learning about using and working alongside AI from an early age. For adults, the report calls on government to invest in skills and training to negate the disruption caused by AI in the jobs market.
Fifth, and aligning with concerns around killer robots, the report says the autonomous power to hurt, destroy or deceive human beings should never be vested in artificial intelligence.
"There is a significant risk that well-intended AI research will be misused in ways which harm people," the report says. "AI researchers and developers must consider the ethical implications of their work".
Advances in AI have the potential to improve outcomes, enhance quality, and reduce costs in such safety-critical areas as healthcare and transportation. Effective and careful applications of pattern recognition, automated decision making, and robotic systems show promise for enhancing the quality of life and preventing thousands of needless deaths.
However, where AI tools are used to supplement or replace human decision-making, we must be sure that they are safe, trustworthy, and aligned with the ethics and preferences of people who are influenced by their actions.
We will pursue studies and best practices around the fielding of AI in safety-critical application areas.
AI has the potential to provide societal value by recognizing patterns and drawing inferences from large amounts of data. Data can be harnessed to develop useful diagnostic systems and recommendation engines, and to support people in making breakthroughs in such areas as biomedicine, public health, safety, criminal justice, education, and sustainability.
While such results promise to provide real benefits, we need to be sensitive to the possibility that there are hidden assumptions and biases in data, and therefore in the systems built from that data - in addition to a wide range of other system choices which can be impacted by biases, assumptions, and limits. This can lead to actions and recommendations that replicate those biases, and have serious blind spots.
Researchers, officials, and the public should be sensitive to these possibilities and we should seek to develop methods that detect and correct those errors and biases, not replicate them. We also need to work to develop systems that can explain the rationale for inferences.
We will pursue opportunities to develop best practices around the development and fielding of fair, explainable, and accountable AI systems.
AI advances will undoubtedly have multiple influences on the distribution of jobs and nature of work. While advances promise to inject great value into the economy, they can also be the source of disruptions as new kinds of work are created and other types of work become less needed due to automation.
Discussions are rising on the best approaches to minimizing potential disruptions, making sure that the fruits of AI advances are widely shared and competition and innovation are encouraged and not stifled. We seek to study and understand best paths forward, and play a role in this discussion.
A promising area of AI is the design of systems that augment the perception, cognition, and problem-solving abilities of people. Examples include the use of AI technologies to help physicians make more timely and accurate diagnoses and assistance provided to drivers of cars to help them to avoid dangerous situations and crashes.
Opportunities for R&D and for the development of best practices on AI-human collaboration include methods that provide people with clarity about the understandings and confidence that AI systems have about situations, means for coordinating human and AI contributions to problem solving, and enabling AI systems to work with people to resolve uncertainties about human goals.
AI advances will touch people and society in numerous ways, including potential influences on privacy, democracy, criminal justice, and human rights. For example, while technologies that personalize information and that assist people with recommendations can provide people with valuable assistance, they could also inadvertently or deliberately manipulate people and influence opinions.
We seek to promote thoughtful collaboration and open dialogue about the potential subtle and salient influences of AI on people and society.
AI offers great potential for promoting the public good, for example in the realms of education, housing, public health, and sustainability. We see great value in collaborating with public and private organizations, including academia, scientific societies, NGOs, social entrepreneurs, and interested private citizens to promote discussions and catalyze efforts to address society's most pressing challenges.
Some of these projects may address deep societal challenges and will be moonshots - ambitious big bets that could have far-reaching impacts. Others may be creative ideas that could quickly produce positive results by harnessing AI advances.
We will assess AI applications in view of the following objectives. We believe that AI should:
The expanded reach of new technologies increasingly touches society as a whole. Advances in AI will have transformative impacts in a wide range of fields, including healthcare, security, energy, transportation, manufacturing, and entertainment. As we consider potential development and uses of AI technologies, we will take into account a broad range of social and economic factors, and will proceed where we believe that the overall likely benefits substantially exceed the foreseeable risks and downsides.
AI also enhances our ability to understand the meaning of content at scale. We will strive to make high-quality and accurate information readily available using AI, while continuing to respect cultural, social, and legal norms in the countries where we operate. And we will continue to thoughtfully evaluate when to make our technologies available on a non-commercial basis.
AI algorithms and datasets can reflect, reinforce, or reduce unfair biases. We recognize that distinguishing fair from unfair biases is not always simple, and differs across cultures and societies. We will seek to avoid unjust impacts on people, particularly those related to sensitive characteristics such as race, ethnicity, gender, nationality, income, sexual orientation, ability, and political or religious belief.
We will continue to develop and apply strong safety and security practices to avoid unintended results that create risks of harm. We will design our AI systems to be appropriately cautious, and seek to develop them in accordance with best practices in AI safety research. In appropriate cases, we will test AI technologies in constrained environments and monitor their operation after deployment.
We will design AI systems that provide appropriate opportunities for feedback, relevant explanations, and appeal. Our AI technologies will be subject to appropriate human direction and control.
We will incorporate our privacy principles in the development and use of our AI technologies. We will give opportunity for notice and consent, encourage architectures with privacy safeguards, and provide appropriate transparency and control over the use of data.
Technological innovation is rooted in the scientific method and a commitment to open inquiry, intellectual rigor, integrity, and collaboration. AI tools have the potential to unlock new realms of scientific research and knowledge in critical domains like biology, chemistry, medicine, and environmental sciences. We aspire to high standards of scientific excellence as we work to progress AI development.
We will work with a range of stakeholders to promote thoughtful leadership in this area, drawing on scientifically rigorous and multidisciplinary approaches. And we will responsibly share AI knowledge by publishing educational materials, best practices, and research that enable more people to develop useful AI applications.
Many technologies have multiple uses. We will work to limit potentially harmful or abusive applications. As we develop and deploy AI technologies, we will evaluate likely uses in light of the following factors:
In addition to the above objectives, we will not design or deploy AI in the following application areas:
We want to be clear that while we are not developing AI for use in weapons, we will continue our work with governments and the military in many other areas. These include cybersecurity, training, military recruitment, veteransâ healthcare, and search and rescue. These collaborations are important and we'll actively look for more ways to augment the critical work of these organizations and keep service members and civilians safe.
Google's announcement was met with immediate scepticism (Newcomer 2018): ""[With the exception of not working on "technologies whose principal purpose or implementation is to cause or directly facilitate injury to people], the rest of the company's "principles" are peppered with lawyerly hedging and vague commitments ... Without promising independent oversight, Google is just putting a new, less persuasive, spin on an old principle it's tried to bury: 'Don't be evil'".
AI designers and developers are responsible for considering AI design, development, decision processes, and outcomes.
1.1. Make company policies clear and accessible to design and development teams from day one so that no one is confused about issues of responsibility or accountability. As an AI designer or developer, it is your responsibility to know.
1.2. Understand where the responsibility of the company/software ends. You may not have control over how data or a tool will be used by a user, client, or other external source.
1,3. Keep detailed records of your design processes and decision making. Determine a strategy for keeping records during the design and development process to encourage best practices and encourage iteration.
1.4. Adhere to your company's business conduct guidelines. Also, understand national and international laws, regulations, and guidelines5 that your AI may have to work within. You can find other related resources in the IEEE Ethically Aligned Design document.
2. Value Alignment
AI should be designed to align with the norms and values of your user group in mind.
2.1. Consider the culture that establishes the value systems you're designing within. Whenever possible, bring in policymakers and academics that can help your team articulate relevant perspectives.
2.2. Work with design researchers to understand and reflect your users' values.
2.3. Consider mapping out your understanding of your users' values and aligning the AI's actions accordingly with an Ethics Canvas. Values will be specific to certain use cases and affected communities. Alignment will allow users to better understand your AI's actions and intents.
AI should be designed for humans to easily perceive, detect, and understand its decision process.
1.1. Allow for questions. A user should b eable to ask why an AI is doing what it's doing on an ongoing basis. This should be clear and up front in the user interface at all times.
1.2. Decision making processes must be reviewable, especially if the AI is working with highly sensitive personal information data like personally identifiable information, protected health information, and/or biometric data.
1.3. When an AI is assisting users with making any highly sensitive decisions, the AI must be able to provide them with a sufficient explanation of recommendations, the data used, and the reasoning behind the recommendations.
1.4. Teams should have and maintain accesstoa record of an AI's decision processes and be amenable to verification of those decision processes.
AI must be designed to minimize bias and promote inclusive representation.
1.1. Real-time analysis of AI brings to light both intentional and unintentional biases. When bias in data becomes apparent, the team must investigate and understand where it originated and how it can be mitigated.
1.2. Design and develop without intentional biases and schedule team reviews to avoid unintentional biases. Unintentional biases can include stereotyping, confirmation bias, and sunk cost bias.
1.3. Instill a feedback mechanism or open dialogue with users to raise awareness of user-identified biases or issues. e.g., Woebot asks ÒLet me know what you think,Ó after suggesting a link.
5. User Data Rights
AI must be designed to protect user data and preserve the user's power over access and uses.
1.1. Users should always maintain control over what data is being used and in what context. They can deny access to personal data that they may find compromising or unfit for an AI to know or use.
1.2. Allow users to deny service or data by having the AI ask for permission before an interaction or providing the option during an interaction. Privacy settings and permissions should be clear, findable, and adjustable.
1.3. Provide full disclosure on how the personal information is being used or shared.
1.4. Users' data should be protected from theft, misuse, or data corruption.
1.5. Forbid use of another company's data without permission when creating a new AI service.
1.6. Recognize and adhere to applicable national and international rights laws when designing for an AI's acceptable user data access permissions.
New developments in Artificial Intelligence are transforming the world, from science and industry to government administration and finance. The rise of AI decision-making also implicates fundamental rights of fairness, accountability, and transparency. Modern data analysis produces significant outcomes that have real life consequences for people in employment, housing, credit, commerce, and criminal sentencing. Many of these techniques are entirely opaque, leaving individuals unaware whether the decisions were accurate, fair, or even about them.
We propose these Universal Guidelines to inform and improve the design and use of AI. The Guidelines are intended to maximize the benefits of AI, to minimize the risk, and to ensure the protection of human rights. These Guidelines should be incorporated into ethical standards, adopted in national law and international agreements, and built into the design of systems. We state clearly that the primary responsibility for AI systems must reside with those institutions that fund, develop, and deploy these systems.
Ethicality of Purpose is driven by the EU Charter of Fundamental Rights:
E1 Beneficence: Do Good
E2 Non maleficence: Do no Harm
E3 Autonomy: Preserve Human Agency
E4 Justice: Be Fair
E5 Explicability: Operate transparently
Achieving Trustworthy AI means that the general and abstract principles need to be mapped into concrete requirements for AI systems and applications. The ten requirements listed below have been derived from the rights, principles and values of Chapter I. While they are all equally important, in different application domains and industries, the specific context needs to be taken into account for further handling thereof.
P0 Perform impact assessment (p.28)
P2 Data Governance
P3 Design for all
P4 Governance of AI Autonomy (Human oversight)
P6 Respect for (& Enhancement of) Human Autonomy [in final section, no. 7]
P7 Respect for Privacy [in final section, no. 6]
1. Supporting Creative Life Styles and Building a Better Society
Through advancing its AI-related R&D and promoting the utilization of AI in a manner harmonized with society, Sony aims to support the exploration of the potential for each individual to empower their lives, and to contribute to enrichment of our culture and push our civilization forward by providing novel and creative types of kando [cf. 0000inspiration]. Sony will engage in sustainable social development and endeavor to utilize the power of AI for contributing to global problem-solving and for the development of a peaceful and sustainable society.
2. Stakeholder Engagement
In order to solve the challenges arising from use of AI while striving for better AI utilization, Sony will seriously consider the interests and concerns of various stakeholders including its customers and creators, and proactively advance a dialogue with related industries, organizations, academic communities and more. For this purpose, Sony will construct the appropriate channels for ensuring that the content and results of these discussions are provided to officers and employees, including researchers and developers, who are involved in the corresponding businesses, as well as for ensuring further engagement with its various stakeholders.
3. Provision of Trusted Products and Services
Sony understands the need for safety when dealing with products and services utilizing AI and will continue to respond to security risks such as unauthorized access. AI systems may utilize statistical or probabilistic methods to achieve results. In the interest of Sony's customers and to maintain their trust, Sony will design whole systems with an awareness of the responsibility associated with the characteristics of such methods.
4. Privacy Protection
Sony, in compliance with laws and regulations as well as applicable internal rules and policies, seeks to enhance the security and protection of customers' personal data acquired via products and services utilizing AI, and build an environment where said personal data is processed in ways that respect the intention and trust of customers.
5. Respect for Fairness
In its utilization of AI, Sony will respect diversity and human rights of its customers and other stakeholders without any discrimination while striving to contribute to the resolution of social problems through its activities in its own and related industries.
6. Pursuit of Transparency
During the planning and design stages for its products and services that utilize AI, Sony will strive to introduce methods of capturing the reasoning behind the decisions made by AI utilized in said products and services. Additionally, it will endeavor to provide intelligible explanations and information to customers about the possible impact of using these products and services.
7. The Evolution of AI and Ongoing Education
People's lives have continuously changed with the advance in technology across history. Sony will be cognizant of the effects and impact of products and services that utilize AI on society and will proactively work to contribute to developing AI to create a better society and foster human talent capable of shaping our collective bright future through R&D and/or utilization of AI.
The eight core principles referred to throughout this report are used as ethical framework to guide organisations in the use or development of AI systems. These principles should be seen as goals that define whether an AI system is operating ethically.
1. Generates net-benefits.
The AI system must generate benefits for people that are greater than the costs.
2. Do no harm.
Civilian AI systems must not be designed to harm or deceive people and should be implemented in ways that minimise any negative outcomes.
3. Regulatory and legal compliance.
The AI system must comply with all relevant international, Australian Local, State/Territory and Federal government obligations, regulations and laws.
4. Privacy protection.
Any system, including AI systems, must ensure people's private data is protected and kept confidential plus prevent data breaches which could cause reputational, psychological, financial, professional or other types of harm to a person.
The development or use of the AI system must not result in unfair discrimination against individuals, communities or groups. This requires particular attention to ensure the Òtraining dataÓ is free from bias or characteristics which may cause the algorithm to behave unfairly.
6. Transparency and explainability.
People must be informed when an algorithm is being used that impacts them and they should be provided with information about what information the algorithm uses to make decisions.
When an algorithm significantly impacts a person there must be an efficient process to allow that person to challenge the use or output of the algorithm.
People and organisations responsible for the creation and implementation of AI algorithms should be identifiable and accountable for the impacts of that algorithm.
Designing AI to be trustworthy requires creating solutions that reflect ethical principles that are deeply rooted in important and timeless values.
AI systems should treat all people fairly
AI systems should empower everyone and engage people
3. Reliability & Safety
AI systems should perform reliably and safely
AI systems should be understandable
5. Privacy & Security
AI systems should be secure and respect privacy
AI systems should have algorithmic accountability
Ethical Principles in the Context of AI Systems (pp.12-13)
E1. Respect for human autonomy
The fundamental rights upon which the EU is founded are directed towards ensuring respect for the freedom and autonomy of human beings. Humans interacting with AI systems must be able to keep full and effective self- determination over themselves, and be able to partake in the democratic process. AI systems should not unjustifiably subordinate, coerce, deceive, manipulate, condition or herd humans. Instead, they should be designed to augment, complement and empower human cognitive, social and cultural skills. The allocation of functions between humans and AI systems should follow human-centric design principles and leave meaningful opportunity for human choice. This means securing human oversight over work processes in AI systems. AI systems may also fundamentally change the work sphere. It should support humans in the working environment, and aim for the creation of meaningful work.
E2. Prevention of harm
AI systems should neither cause nor exacerbate harm or otherwise adversely affect human beings. This entails the protection of human dignity as well as mental and physical integrity. AI systems and the environments in which they operate must be safe and secure. They must be technically robust and it should be ensured that they are not open to malicious use. Vulnerable persons should receive greater attention and be included in the development, deployment and use of AI systems. Particular attention must also be paid to situations where AI systems can cause or exacerbate adverse impacts due to asymmetries of power or information, such as between employers and employees, businesses and consumers or governments and citizens. Preventing harm also entails consideration of the natural environment and all living beings.
The development, deployment and use of AI systems must be fair. While we acknowledge that there are many different interpretations of fairness, we believe that fairness has both a substantive and a procedural dimension. The substantive dimension implies a commitment to: ensuring equal and just distribution of both benefits and costs, and ensuring that individuals and groups are free from unfair bias, discrimination and stigmatisation. If unfairbiases can be avoided, AI systems could even increase societal fairness. Equal opportunity in terms of access to education, goods, services and technology should also be fostered. Moreover, the use of AI systems should never lead to people being deceived or unjustifiably impaired in their freedom of choice. Additionally, fairness implies that AI practitioners should respect the principle of proportionality between means and ends,and consider carefully how to balance competing interests and objectives. The procedural dimension of fairness entails the ability to contest and seek effective redress against decisions made by AI systems and by the humans operating them. In order to do so, the entity accountable for the decision must be identifiable, and the decision-making processes should be explicable.
Explicability is crucial for building and maintaining users' trust in AI systems. This means that processes need to be transparent, the capabilities and purpose of AI systems openly communicated, and decisions - to the extent possible - explainable to those directly and indirectly affected. Without such information, a decision cannot be duly contested. An explanation as to why a model has generated a particular output or decision (and what combination of input factors contributed to that) is not always possible. These cases are referred to as 'blackbox' algorithms and require special attention. In those circumstances, other explicability measures (e.g. traceability, auditability and transparent communication on system capabilities) may be required, provided that the system as a whole respects fundamental rights. The degree to which explicability is needed is highly dependent on the context and the severity of the consequences if that output is erroneous or otherwise inaccurate.
Requirements of Trustworthy AI (pp.14-20, 26-31):
R1 Human agency and oversight (pp.15-16, 26-27)
Including fundamental rights, human agency and human oversight
AI systems should ... act as enablers to a democratic, flourishing and
equitable society by supporting the user's agency and foster fundamental
where ... risks exist, a fundamental rights impact assessment should be undertaken.
... mechanisms should be put into place to receive external feedback regarding AI systems that potentially infringe on fundamental rights.
... the right not to be subject to a decision based solely on automated processing when this produces legal effects on users or similarly significantly affects them.
Human oversight ... may be achieved through governance mechanisms such as a human-in-the- loop (HITL), human-on-the-loop (HOTL), or human-in-command (HIC) approach.
... the less oversight a human can exercise over an AI system, the more extensive testing and stricter governance is required.
R2 Technical robustness and safety (pp.16-17, 27-28)
Including resilience to attack and security, fall back plan and general safety, accuracy, reliability and reproducibility
A crucial component of achieving Trustworthy AI is technical robustness.
... requires that AI systems be developed with a preventative approach to risks and in a manner such that they reliably behave as intended while minimising unintentional and unexpected harm, and preventing unacceptable harm.
... protected against vulnerabilities ...
... steps should be taken to prevent and mitigate ... [unintended applications ... and potential abuse] ...
Resilience ... AI systems should have safeguards that enable a fallback plan in case of problems.
... it is crucial for safety measures to be developed and tested proactively.
It is critical that the results of AI systems are reproducible ... Reproducibility describes whether an AI experiment exhibits the same behaviour when repeated under the same conditions.
R3 Privacy and data governance (pp.17, 28)
Including respect for privacy, quality and integrity of data, and access to data
Prevention of harm to privacy also necessitates adequate data governance that covers the quality and integrity of the data used, its relevance in light of the domain in which the AI systems will be deployed, its access protocols and the capability to process data in a manner that protects privacy.
[The risk that data contains] socially constructed biases, inaccuracies, errors and mistakes ... needs to be addressed prior to training with any given data set.
Processes and data sets used must be tested and documented at each step such as planning, training, testing and deployment.
R4 Transparency (pp.18, 28-29)
Including traceability, explainability and communication
The data sets and the processes [and the decisions made by the AI system] ... should be documented to the best possible standard to allow for traceability ...
explanations of the degree to which an AI system influences and shapes the organisational decision-making process, design choices of the system, and the rationale for deploying it, should be available ...
Traceability facilitates auditability as well as explainability.
Technical explainability requires that the decisions made by an AI system can be understood and traced by human beings.
... explanation should be timely and adapted to the expertise of the stakeholder concerned ...
... humans have the right to be informed that they are interacting with an AI system.
... the option to decide against this interaction in favour of human interaction should be provided where needed to ensure compliance with fundamental rights.
R5 Diversity, non-discrimination and fairness (pp.18-19, 29-30)
Including the avoidance of unfair bias, accessibility and universal design, and stakeholder participation
... systems should be ... designed in a way that allows all people to use AI products or services, regardless of their age, gender, abilities or characteristics.
... consult stakeholders who may directly or indirectly be affected by the system throughout its life cycle ... ensuring ... information, consultation and participation throughout the whole process
... solicit regular feedback even after deployment
R6 Societal and environmental wellbeing (pp.19, 30-31)
Including sustainability and environmental friendliness, social impact, society and democracy
... the broader society, other sentient beings and the environment should be also considered as stakeholders throughout the AI system's life cycle.
Beyond assessing the impact of an AI system's development, deployment and use on individuals, this impact should also be assessed from a societal perspective, taking into account its effect on institutions, democracy and society at large ... including not only political decision-making but also electoral contexts.
R7 Accountability (pp.19-20, 31)
Including auditability, minimisation and reporting of negative impact, trade-offs and redress.
... mechanisms [must] be put in place to ensure responsibility and accountability for AI systems and their outcomes, both before and after their development, deployment and use.
AI systems should be able to be independently audited .... Auditability entails the enablement of the assessment of algorithms, data and design processes.
The use of impact assessments ... both prior to and during the development, deployment and use of AI systems can be helpful to minimise negative impact. These assessments must be proportionate to the risk that the AI systems pose.
... trade-offs should be explicitly acknowledged and evaluated ... , reasoned and properly documented ...
In situations in which no ethically acceptable trade-offs can be identified, the development, deployment and use of the AI system should not proceed in that form.
The decision-maker must be accountable for the manner in which the appropriate trade-off is being made, and should continually review the appropriateness of the resulting decision to ensure that necessary changes can be made to the system where needed.
When unjust adverse impact occurs, accessible mechanisms should be foreseen that ensure adequate redress.
ACM (2017) 'Statement on Algorithmic Transparency and Accountability' Association for Computing Machinery, January 2017, at https://www.acm.org/binaries/content/assets/public-policy/2017_usacm_statement_algorithms.pdf
Asimov I. (1942) 'Runaround' (originally published in 1942), reprinted in Asimov I. 'I, Robot' Grafton Books, London, 1968, pp. 33- 51
BS (2016) 'Robots and robotic devices. Guide to the ethical design and application of robots and robotic systems' British Standards Institute, 2016
Clarke R. (1993) 'Asimov's Laws of Robotics: Implications for Information Technology' In two parts, in IEEE Computer 26,12 (December 1993) 53-61, and 27,1 (January 1994) 57-66, at http://www.rogerclarke.com/SOS/Asimov.html
CLA-EP (2016) 'Recommendations on Civil Law Rules on Robotics' Committee on Legal Affairs of the European Parliament, 31 May 2016, at http://www.europarl.europa.eu/sides/getDoc.do?pubRef=-//EP//NONSGML%2BCOMPARL%2BPE-582.443%2B01%2BDOC%2BPDF%2BV0//EN
CSIRO (2019) 'Artificial Intelligence: Australia's Ethics Framework: A Discussion Paper' CSIRO, April 2019, at https://consult.industry.gov.au/strategic-policy/artificial-intelligence-ethics-framework/supporting_documents/ArtificialIntelligenceethicsframeworkdiscussionpaper.pdf
Devlin H. (2016). 'Do no harm, don't discriminate: official guidance issued on robot ethics' The Guardian, 18 Sep 2016, at https://www.theguardian.com/technology/2016/sep/18/official-guidance-robot-ethics-british-standards-institute
EC (2018) 'Draft Ethics Guidelines for Trustworthy AI' High-Level Expert Group on Artificial Intelligence, European Commission, 18 December 2018, at https://ec.europa.eu/newsroom/dae/document.cfm?doc_id=57112
EC (2019) 'Ethics Guidelines for Trustworthy AI' High-Level Expert Group on Artificial Intelligence, European Commission, April 2019, at https://ec.europa.eu/newsroom/dae/document.cfm?doc_id=58477
FLI (2017) 'Asilomar AI Principles' Future of Life Institute, January 2017, at https://futureoflife.org/ai-principles/?cn-reloaded=1
GEFA (2016) 'Position on Robotics and AI' The Greens / European Free Alliance Digital Working Group, November 2016, at https://juliareda.eu/wp-content/uploads/2017/02/Green-Digital-Working-Group-Position-on-Robotics-and-Artificial-Intelligence-2016-11-22.pdf
Google (2018) 'Objectives for AI applications' Google, June 2018, at https://www.blog.google/technology/ai/ai-principles/
Hirano (2017) 'AI R&D guidelines' Proc. OECD Conf. on AI developments and applications, October 2017, http://www.oecd.org/going-digital/ai-intelligent-machines-smart-policies/conference-agenda/ai-intelligent-machines-smart-policies-hirano.pdf
HOL (2018) 'AI in the UK: ready, willing and able?' Select Committee on Artificial Intelligence, House of Lords, April 2018, at https://publications.parliament.uk/pa/ld201719/ldselect/ldai/100/100.pdf
IBM (2018) 'Everyday Ethics for Artificial Intelligence' IBM, September 2018, at https://www.ibm.com/watson/assets/duo/pdf/everydayethics.pdf
IEEE (2017) 'Ethically Aligned Design', Version 2. IEEE, December 2017. at http://standards.ieee.org/develop/indconn/ec/autonomous_systems.html
ISOC (2017) 'Artificial Intelligence and Machine Learning: Policy Paper' Internet Society, April 2017, at https://www.internetsociety.org/resources/doc/2017/artificial-intelligence-and-machine-learning-policy-paper/
ITIC (2017) 'AI Policy Principles' Information Technology Industry Council, undated but apparently of October 2017, at https://www.itic.org/resources/AI-Policy-Principles-FullReport2.pdf
JSAI (2017) 'Ethical Guidelines' The Japanese Society for Artificial Intelligence, May 2017, at http://ai-elsi.org/wp-content/uploads/2017/05/JSAI-Ethical-Guidelines-1.pdf
MS (2019) 'Microsoft AI Principles' Microsoft, undated but apparently of April 2019, at https://www.microsoft.com/en-us/ai/our-approach-to-ai
Newcomer E. (2018). 'What Google's AI Principles Left Out: We're in a golden age for hollow corporate statements sold as high-minded ethical treatises' Bloomberg, 8 June 2018, at https://www.bloomberg.com/news/articles/2018-06-08/what-google-s-ai-principles-left-out
Pichai S. (2018) 'AI at Google: our principles' Google Blog, 7 Jun 2018, at https://www.blog.google/technology/ai/ai-principles/
PoAI (2018) 'Our Work (Thematic Pillars)' Partnership on AI, April 2018, at https://www.partnershiponai.org/about/#pillar-1
Smith R. (2018). '5 core principles to keep AI ethical'. World Economic Forum, 19 Apr 2018, at https://www.weforum.org/agenda/2018/04/keep-calm-and-make-ai-ethical/
Sony (2019) ' Sony Group AI Ethics Guidelines' Sony, 1 Mar 2019, at https://www.sony.net/SonyInfo/csr_report/humanrights/hkrfmg0000007rtj-att/AI_Engagement_within_Sony_Group.pdf
TPV (2018) 'Universal Guidelines for Artificial Intelligence' The Public Voice, October 2018, at https://thepublicvoice.org/ai-universal-guidelines/
UGU (2017) 'Top 10 Principles for Ethical AI' UNI Global Union, December 2017, at http://www.thefutureworldofwork.org/media/35420/uni_ethical_ai.pdf
Roger Clarke is Principal of Xamax Consultancy Pty Ltd, Canberra. He is also a Visiting Professor in Cyberspace Law & Policy at the University of N.S.W., and a Visiting Professor in the Research School of Computer Science at the Australian National University. He has also spent many years on the Board of the Australian Privacy Foundation, and is Company Secretary of the Internet Society of Australia.
The content and infrastructure for these community service pages are provided by Roger Clarke through his consultancy company, Xamax.
From the site's beginnings in August 1994 until February 2009, the infrastructure was provided by the Australian National University. During that time, the site accumulated close to 30 million hits. It passed 65 million in early 2021.
Sponsored by the Gallery, Bunhybee Grasslands, the extended Clarke Family, Knights of the Spatchcock and their drummer
Xamax Consultancy Pty Ltd
ACN: 002 360 456
78 Sidaway St, Chapman ACT 2611 AUSTRALIA
Tel: +61 2 6288 6916
Created: 11 July 2018 - Last Amended: 26 August 2020 by Roger Clarke - Site Last Verified: 15 February 2009
This document is at www.rogerclarke.com/EC/GAIP.html