Bias in AI is a key topic of concern

In News

• UNESCO’s framework for moral AI would have a far-reaching influence on the complete gamut of AI actions.
  • The facets of AI which will be affected are
    • improvement, 
    • software, 
    • ethicality, 
    • information privateness, and 
    • regulation.
  • It is still under consultation and not fully developed.

UNESCO’s Framework for Moral AI

• It is an evolving framework encompassing all stakeholders and tries to address challenges faced by users due to systematic flaws.
• Aim of this Framework: 
  • To have a holistic and evolving framework of values, ideas and actions 
  • To inform societies in dealing responsibly, with the recognized and unknown influence of AI on human beings and society at massive. 
  • To provide a moral framework to judge the AI across the world
  • To constructing neutral, non-discriminative and delicate AI platforms
• UNESCO’s framework for moral AI can have a far-reaching influence within the area of AI like
  • sensitivity to privateness and inclusion; 
  • transparency, equity and non-discriminant; 
  • accountability by way of participation; 
  • mindset change that helps a sustainable AI setting and a correct steadiness between enterprise development and promotion of human values.

Need of a Universal Framework for AI Ethics

• Challenge of Biased data for training AI:
  • AI methods, we all know, are in-built and try to mimic human intelligence. 
  • The methods are taught by way of sharing monumental information of previous human actions to be taught from.
  • The major problem is that the data and historic information are studded with biases and discrimination.
  • It may be catastrophic as humans have certain limited capability but an ill trained AI can leave a significant impact on worldwide infrastructure.
• Chances of Racial Discrimination:
  • Further, a bias against people of colour was found in AI of a criminal tracking network, which led to multiple misjudgments and false arrests.
• Gender Bias:
  • Different studies have been found to inherently promote gender bias.
  • So, the AI too may pick up these lines of thinking creating a broken unreliable system.
• Mode of Consent/ Right to Privacy:
  • With extra data about individuals and sharper algorithms, better suggestions, therapies and services are possible.
  • But it can also create a moral hazard.
  • Too much data can be used by notorious elements to harm a person or his integrity and dignity.
  • It is against the much debated right to privacy.
  • Further, shall the pop up on Windows be considered as true consent as most users without understanding the consequences, choose to give consent through pop ups.
  • The best example is the Cambridge Analytica Case where the choice of US Citizens were manipulated using data collected by Facebook.
• Lack of enough Knowledge and Sensitiveness among Developers:
  • It is a major bottleneck as the developers, themselves, know very less about their products and ethical concerns associated with it.
  • The need is to involve different stakeholders on the development side of AI rather than only focusing on users.
• Effect on the users of AI:
  • The newer algorithms of Google and Facebook are promoting polarisation by showing people only what they want to see.
  • It is creating a less tolerant society which lives in its own bubble.

Solutions/ Way Forward suggested by Ethical Code by UNESCO

• Systematic Method to access the biases and cure them:
  • To eliminate Gender and other biases, a systematic way of the AI improvement lifecycle needs to be developed. 
    • This system should include all dimensions including problems to prove that it has certainly been taken care of.
  • Accountability and possession need to be well-defined.
    • It will make sure equity and inclusion are adopted proper by way of the event, validation, and supply of an AI system. 
    • Extra importantly, all these components need to be a part of a regulatory and compliance framework.
  • Information assortment approaches and information distribution need to be streamlined to keep away from bias.
• Role of Developers and Management:
  • Builders ought to make sure that the scope is neither myopic nor meant to usher in any intersection disparity. 
  • Moreover, the developers and management should retain the general societal influence of the outcomes in view.
  • The parameters which will play a really key function to make sure equity and moral finish in the final result are 
    • Information collections
    • Data curation and 
    • The methodology concerned 
• More in depth studies on effect of AI on our life:
  • On this journey, simply consciousness just isn’t sufficient however recognition, acceptance and addressing of the difficulty is extra vital. 
  • The frameworks that assist a holistic inclusion of coders, builders, and decision-makers must be authorised and promoted. 
• Promoting Stakeholder Partnership:
  • Seeing the massive impact of internet giants on a person’s behaviour, Search Engines need to balance between suitable search outputs and unbiased, conflict free and secular contents.

Conclusion

• Technology needs to serve humanity not the vice versa.
• A technology without ethics will be nothing more than a ticking time bomb.
• There is an urgent need to tackle the moral and ethical issues related to AI through an universal code of ethics which emcompasses all.

Source: TH