WHAT IS AI?
Artificial intelligence is a computerised software which simulates human decision-making and cognisance to perform tasks. In order to produce such human-like output it needs to consume a large amount of data so that it can learn to recognise patterns, understand the priority factors, and analyse and replicate the previous decisions or tasks. Now-a-days repetitive and monotonous tasks can be delegated to the AI system to be performed through a procedure called the Robotic Process Automation. It reduces the burden on human beings and frees the workforce to devote their time towards ideological advancements rather than performing simple, menial, and tedious work. For instance, machine learning and artificial intelligence can be programmed to predict outcomes of cases by providing them data on facts of the present case, judges involved, previous judgments, etc. The system may then sort out the facts of the present case, analyse the political linings of the judges or their general tendencies towards topics involved in the case and recognise patterns in the precedents to come to a logical conclusion.
THE ETHICAL DILEMMA
‘There has been an abominable deterioration of ethical standards primarily due to mechanisation and depersonalisation of our lives, which is a disastrous by-product of science and technology’, says Albert Einstein himself.
Artificial intelligence has undoubtedly created opportunities around the globe. It has revolutionised the healthcare industry by providing automated patient care and digital aids in hospitals and clinics. Article 24 of the Universal Declaration on Human Rights says that ‘Everyone has the right to rest and leisure, including reasonable limitation of working hours and periodic holidays with pay.’ AI has made it possible through the invention of robots, which are a consequence of the merging of artificial intelligence and machines so as to provide AI a medium to interact with the physical world, which can now perform dangerous tasks in work environment without any time constraints.
Every technological advancement has a drawback, it is imperative that such a red flag be discovered and addressed for the said advancement to be truly fruitful. Then, what shall be the flaw of the rising Artificial Intelligence systems? It is undoubtedly its ignorance of human rights and neglect for ethics.
PRIVACY CONCERNS
Let’s go back to the example of the healthcare systems where AI has modernised its functioning and implementation. An AI system will acquire the patient history and other general data in the process of providing care to the patients and performing the administrative tasks. This is crucial for its proper working since it uses such data for configurations. But the data collected will involve demographic information, patient records or treatment methods which the patient might not be willing to share with others and since the AI system then becomes a store of information, it becomes a target for external attacks. Such sensitive medical information then has a possibility of becoming exposed to the public which can be further utilised for various ill-purposes such as human trafficking.
Therefore, it can be stated that AI can pose harmful security concerns. Such privacy breaches can be further highlighted in cases where the use of such softwares leads to unnecessary risks of surveillance. For instance, Chicago is the one of the most surveilled cities in the world due to the vast number of CCTVs, their tight integration, and their capacity to gather and analyse information. Such a security infrastructure gives the government the power to monitor anyone in public spaces. Chicago’s surveillance policies can be regarded as pervasive and poorly regulated in the themes of human rights and privacy. It further raises concerns regarding the future possibilities and advancements if the AI is improvised and implemented in various walks of life around the world.
APPARENT BIAS IN DATA PROCESSING
It has been made abundantly clear before that an AI system needs an abundance of data to perform its tasks, the data collected by it from its surroundings or the data fed to it manually may involve information related to gender, caste, race, religion, demography, linguistic preferences etc. of the person concerned. The question that arises here is whether it is possible to avoid vulnerability, discrimination, and exclusion when such data is made available to the AI system. Many might make the claim that as a computer system, such bias is not inherent in its behaviour. But let me remind you once again that artificial intelligence softwares are unlike other computational systems, it imitates human actions and takes inspiration from decisions taken in the past. Even if biases are not apparent in its functioning, there is a high chance that the biases of the developer may have seeped into the code resulting in a lopsided output. Also, past discriminations which have been recorded and which is accessed by the AI system to modify its own outcomes, may influence the decision-making process. Such past discriminations hidden in the data which could be identified through pattern recognition by AI may result in algorithmic discrimination.
In the 2018 Gender Shades Project, Joy Buolamwini and Timnit Gebru tested the image-processing algorithm developed by Microsoft, IBM, and Face++ in making classifications based on skin colour and gender. All three algorithms performed worse on darker-skinned females and showed an error rate of 34% higher than that for lighter-skinned males. Such discrepancies are concerning, and the companies need to ensure that the services they provide are equitable, both in terms of technology and its application.
CONVENTION 108
Convention for the Protection of Individuals with regard to Automatic Processing of Personal Data, also known as Convention 108 is the first international instrument for data protection. It was opened for signature on 28 January 1981 by the Council of European Union. The transboundary flows of personal data undergoing automated data processing has increased and has become easier to perpetuate. The convention aims to secure the individuals’ right to privacy. Article 4of the Convention elaborates on the duties of the signatory parties to take necessary steps in the domestic laws of the respective countries to give effect to the basic principles outlined in the chapter, and it has been enforced that such measures should be taken latest at the time of entry into the Convention by that Party. This positively promotes the signatory countries to make changes to their legislations, instead of providing them a timeframe for policy actions which at most times is ineffective. Article 25 further rejects the proposal for reservations in regard to any of the provisions of the Convention. Article 7 talks about data security and makes it mandatory for the concerned authorities to take security measures to prevent unauthorised access, alteration, or dissemination of personal data.
The General Data Protection Regulation (GDPR) which sets out rules and regulations for businesses worldwide to process personal data, replaced the EU’s Data Protection Directive in 2018. In 2022, GDPR was used to regulate AI in Italy where a company, Clearview AI stored biometrics and geolocation data without any legal basis and was consequently fined with €20 million. It is a common consensus that the fines imposed by GDPR are hefty if it discovers any breach of regulation. If the breach is significant as listed in Article 83(5) of the GDPR a maximum penalty of €20 million or 4% of their annual global turnover, whichever is higher is imposed. For less severe infractions, as listed in Article 83(4) of the GDPR, €10 million or up to 2% of their annual global turnover is imposed.
Since the data in question is personal and hence sensitive, such stringent regulations are appreciated and it helps the world community to inch a step closer towards data privacy and human rights.
CONCLUSION
The complexity of ethical issues surrounding AI requires the involvement of multiple stakeholders from various levels and sectors of the international community. Firstly, the AI technology shall be modified as such that models of social justice and fairness are upheld. This can be primarily done through evaluation of the quality of training data for the AI systems. Proper risk assessments shall be performed before employing artificial intelligence, so that there is a limit to its applicability and such limitations shall be set after cost-benefit analysis. The AI systems must also be auditable and traceable, and the final responsibility shall lie with humans, the AI shall not displace humans in terms of responsibility and accountability. Artificial intelligence has posed some challenges to securing human rights in the global scenario, but we must view this challenge as crucial to the process of development and find solutions to overcome it thereby allowing for human rights promotion alongside technological advancements.
Author: Srisoniya Subramoniam
University and Year (if applicable): O.P. Jindal Global University/ 1st year
Programme: BA LLB (Hons.)
LinkedIn Page Link: https://www.linkedin.com/in/srisoniya