Importance of Data Ethics in an AI-Driven World

Understanding AI

“Artificial intelligence” (AI) has transformed the way we live, interact, and do business. We live in the times of “augmented intelligence”, “ubiquitous sensing systems” and “embedded technologies” and “big data”. The internet of things, use of Siri, Alexa, virtual chatbots and smart devices have become a norm. As per reliable sources, the global AI market in 2016 was around USD 260 billion and it is estimated to exceed 3 trillion by 2024¹. While AI brings multiple efficiencies of time and productivity, it is also prone to misuse by mankind. Therefore, in today’s time the importance of AI, data science and data ethics cannot be undermined.

The human mind does complex problem solving and humans also have the ability to create machines capable of replicating its strategic thinking and ingrained processes. AI is the science of deciphering and decoding how a human brain thinks, works and how it does problem-solving. AI is different from robotics which mainly concentrates on automating manual tasks. However, AI performs high-volume complex tasks requiring human life intelligence effortlessly. Although, the overarching goal of AI is to build software that can create (and even surpass) efficiencies of human intelligence, AI uses machine learning as AI does not have an IQ of its own. Another distinction is that AI is incapable of feeling, of being able to form any intention and cannot possibly have beliefs but operates via simulations and commands. John McCarthy described “artificial intelligence” as “the science and engineering of making intelligent machines, especially intelligent computer programs”². He did not describe it with words that connote it can feel an emotion or form an intention. For instance, Microsoft launched Robot Tay on social media, but it did not succeed in recognising positive and negative expressions of human emotions as it lacked emotional intelligence. In the author’s view, artificial intelligence can be defined as “the study of human intelligence successfully simulated through algorithms and programs to produce predictive and logical results equal to or superior to human intelligence for certain specified tasks”. In 1997, “Deep Blue”, IBM supercomputer had beaten the then world chess champion, Garry Kasparov in chess in a rematch after its initial defeat in 1996. This is an apt example of artificial intelligence trumping human intelligence in early times³. Apple is reported to work on a chip for AI known as the “Apple Neural Engine” which uses speech and facial recognition, biometrics, and other technologies. Siri was released in 2011. Amazon has also created Alexa which is similar to the Siri application⁴. Facebook is using AI to curb fake news online on their platforms.⁵ Driverless cars and advances in science using AI for treating diseases are all AI-based⁶.ChatGPT, an AI recently released by a firm called OpenAI, based in San Francisco. “GPT” is an abbreviation for Generative Pre-Trained Transformer. It is an AI chatbot that assists an individual with various tasks while having human-like conversations. It employs a machine learning technique called “Reinforcement Learning from Human Feedback” (RLHF) that has been trained to respond, recognise, and react to specific patterns. ChatGPT can be used to simulate dialogue, answer follow-up questions, challenge incorrect premises and reject inappropriate requests, challenge a false premise in a question, respond to questions and answers, compose texts, translate between languages, summarise the text and detect keywords.

Subfields of AI

AI includes in its ambit the following subfields of science—

(i) Machine learning. — Machine learning uses statistics, physics, neural networks, and research to make inferences from the data and learn the inferential skill over time.⁷

(ii) A neural network is a kind of machine learning comprising of interconnected units like neurons which process information in response to external stimuli called inputs which send information between each unit. The process studies multiple passes exchanged between the connections to inferred meaning from undefined data.

(iii) Deep learning uses large number of neural networks with multiple layers of processing of units based on advanced computing technologies to understand complex patterns in big data. Few popular applications include image and speech recognition systems.

(iv) Cognitive computing is also a subfield of AI that aims at creating human-like interaction with machines. It works on simulating human processes through its capability of interpreting images and speech and making a response which is logical.

(v) Computer vision is another field of AI using pattern recognition and deep learning to understand what is shown in the picture or a video. The machines capture images or videos in real-time and are used to interpret them and the surroundings shown therein.

(vi) NLP in other words natural language processing connotes the power of computers to analyse and generate human language such as speech. Another level of NLP is natural language interaction which facilitates humans to communicate with computers using ordinary language (or conversational AI) to perform complex tasks such as use of AI-based robots in restaurants to serve its customers.

Need for data ethics in AI and data science

As AI is being leveraged by nations across various sectors of industry, their regulatory regime applicable to AI is also transforming rapidly, inter alia, creation of new data protection laws or adapting existing regulations, codifying data ethics and data science have become indispensable in present AI times.⁸ This will ensure development and deployment of AI is both responsible and facilitates sustainable growth with due regard to principles of natural justice, fairness, transparency, and accountability, in letter and spirit.

Avoid black box scenario

With use of AI and machine learning, AI will make shortcuts for decision-making creating black box scenario. It is thus essential that engineers who created an AI system can explain the choices it has made to make its decisions/inferences. It is therefore important to cull out a best practice to ensure innovation is responsible innovation and not otherwise. In order to create a sustainable AI technology, it is imperative to codify basic principles of data protection and ethical data handling in data science.⁹ “Data science” means scientific study of creation, manipulation, and transformation of data to create meaning.¹⁰ Some of the key principles of data ethics include confidentiality, non-discrimination, requirement of consent, transparency, and non-disclosure.¹¹ Use of anonymisation, de-identification, pseudonymisation, in data handling processes and compliance with applicable laws such as GDPR/personal data protection laws. Articles 13 to 15 of the General Data Protection Regulation (GDPR)¹² stipulate data subject has right to access meaningful information about logic involved, significance and envisaged consequences of automated decision-making systems such as profiling.

By way of an example, Amazon used algorithms for many years to make initial hiring filters and built an internal algorithm for hiring purposes. The algorithm over time became highly sexist. As more men applied for a job at Amazon, through machine learning Amazon’s computers taught itself that male candidates are preferred. The glitch was observed and later rectified. Recently, Idaho passed a law specifying methods and data used in bail algorithms must be made publicly available on account of maintaining transparency and accountability.¹³ With reason codes, data robots can explain any automated decision from machine learning models. Model x-ray provides a way to demonstrate the behaviour of a model to stakeholders and regulators. In India, the proposed Digital India Act, 2023¹⁴ also prescribes provisions that address artificial intelligence and risks and reiterates the need for the deployment of responsible and ethical use of AI and emerging technologies.

Data protection

In order to prevent discrimination, GDPR disallows companies from using the processing of personal data revealing racial or ethnic origin, political opinions, religious or philosophical beliefs, or trade union memberships, processing of genetic data, biometric data for uniquely identifying a natural person, health data or person’s sex life or sexual orientation.¹⁵

Though GDPR has laid down a clear blueprint for how data protection is addressed in Europe or for companies serving European customers, India is yet to have a comprehensible law on data protection and privacy. In K.S Puttaswamy v. Union of India¹⁶, the nine-Judge Bench of the Supreme Court of India (commonly known as the “Aadhaar judgment”), held that right to privacy is a fundamental right of every citizen of India. The Ministry of Electronics and Information Technology (MeitY) has released a new Data Protection Bill on 18-11-2022 i.e. three months after striking down a prior version which was named the Personal Data Protection Bill, 2019¹⁷. Salient principles of prior explicit consent, accountability, transparency, retention as is necessary, confidentiality, right to object, right of updation, deletion of content, find its footprint also in the recently issued the Digital Personal Data Protection Bill, 2022.¹⁸ This Bill adopts a consent-based approach which is also adopted in European Union’s GDPR law. It includes provisions on “purpose limitations” for data collection, reasons for collecting and processing personal data, rectifying errors in data or, withdrawal of consent by a user, easing of cross-border data flows, and imposes hefty penalties on organisations up to 500 crores that violate the Bill’s data protection requirements. The Bill also provides for deemed consent by a user such as in cases of medical emergency or compliance with a judgment or order of a court. Therefore, nations will require creation of new laws or adaptation of their existing laws to strengthen their data protection regimes and codify data ethics (to the extent possible as proposed also within EU) as part of their regulatory regime for responsible use of AI and future technologies.¹⁹ Courts in different parts of the world have also begun to deploy AI to enhance efficiency of their justice delivery systems. For instance, in America as well as India, AI-based robots have been deployed to decide traffic challan cases. In Canada, AI is being deployed in courts to replace mundane administrative tasks. In India, AI can be leveraged to enhance access to justice and justice delivery system. Justice D.Y. Chandrachud, the current Chief Justice of India has envisioned and directed measures for the digitisation and adoption of emerging tech in the Indian judicial system. Data ethics by design will be quintessential to AI’s successful adoption in India, particularly, in the justice delivery system.

Cybersecurity and AI

AI has been deployed by many countries such as US and China for national security and in armed forces for more than a decade. Robot snipers, drones, and turrets have been developed for national defence purposes. AI-based intelligent weaponry has been created/used by many nations so far. However, there is a need for developing new regulations, conventions, and international treaties to define a set of rules for the use of AI in warfare. India is currently in midst of formulation of the national cybersecurity policy for the country which is likely to address this aspect as well. A comprehensive report “Credible Cyber Deterrence in Armed Forces of India” was released by Vivekananda International Foundation (VIF) in 2019 which discusses various changes needed in the Indian cybersecurity and legal framework to address cyber warfare, AI-based weaponry and need to form international arrangements to address these serious concerns.²⁰ Where AI is used to commit cyber terrorism, Section 66-F of the extant Information Technology Act, 2000²¹ will apply. It prescribes cyber terrorism as an offence punishable with punishment up to life imprisonment. This provision will apply in case AI-enabled arms and technology are used to commit cyberterrorism, hack sensitive data, or affect critical information infrastructure of the country. The US military has just released a prospective document entitled “Robotic and Autonomous Systems Strategy.” It lists certain objectives therein, namely, increase knowledge abilities in operations, improving logistics capacity; facilitation of movement and manoeuvring; increase the protection of forces.²² India is also in the process of formulating detailed action plan to address AI and national cybersecurity and cyber defence strategy and related regulations. NITI Aayog has issued strategic action plan for development and deployment of AI in India. The NITI Aayog in its National Strategy for AI in 2018²³ laid down a framework for aggregating and disseminating knowledge of AI through concerted nationwide and sector-wise initiatives, inter alia, suggesting establishment of centres of excellence and building collaborative framework for research and development, skilling, re-skilling and adoption and implementation of AI in India. In its Report issued in 2021 on responsible use of AI²⁴, NITI Aayog recommended setting up of independent legal advisory body named Centre for Ethics and Technology (CET). CET was envisioned to serve as a think tank with multi-stakeholder representation and develop guidelines or ethics review mechanisms to evaluate efficacy of AI systems developed and used in India.

Similarly, across other nations appropriate regulations will need to be framed to deal with AI-based cyberattacks, cyberwarfare using AI-enabled virtual bots and malicious software. Can virtual bots be imputed criminal intention and be criminally liable? Do they possess a legal personality? Also, how will law address a situation if control over an AI-enabled device is lost? These are some pressing legal issues that require serious deliberation globally.

The juxtaposition of AI, cybersecurity and data ethics is pivotal to responsible deployment of this technology for the world to harness its optimal potential. Its ethical use lies in leveraging it in a sustainable manner for the benefit of all mankind. This aligns with the ethos of the India’s presidency in G20 and echoes the sentiment of “One Earth One Family One Future.” AI has an invaluable role to drive innovation and progress across all sectors of activity. It is imperative to deliberate on its responsible use to channelise its unique potential leading to sustainable growth. On this premise alone, can we act in the best interest of all and together build a secure future for the world.