Rationale for high-stakes AI decisions ‘must be public and transparent’

Artificial Intelligence (AI) must be demystified in order for the Irish public to embrace it, according to a new strategy document that  set out a series of steps to ensure widespread societal buy-in.

The Government document also predicts that one in three jobs in Ireland is likely to be disrupted by AI, though some will involve changes to roles and tasks, rather than actual job losses. 

The OECD has also estimated that increasing use of AI in the public sector will free up nearly one-third of public servants’ time.

Inevitably, computers will complement or replace specific tasks otherwise performed by people, the document says.

Upskill

The Government will prioritise policies that ensure workers can upskill or reskill, the strategy document continues.

And other interventions, including in the labour-market regulation and social-protection spheres, will be needed in order to protect workers’ interests, it adds.

Within the justice sector, AI has considerable implications for ethics, human rights and the rule of law, the document points out.

The Council of Europe’s ethical charter on the use of AI in judicial systems identifies suitable low-risk areas:

Development of case law by linking doctrine, case law, laws and regulations,

• Determination of compensation awards,
• Online dispute resolution.

Artificial Intelligence (AI) refers to machine-based systems, with varying levels of autonomy, that can use data for human-defined objectives to make predictions, recommendations or decisions.

Artificial intelligence (AI) will be at the forefront of the transition to a greener, more sustainable, and more digital, economy, according to the foreword by Robert Troy (Minister of State for Trade Promotion, Digital and Company Regulation).

The document sets out the following steps:

• AI ambassador to champion AI as a positive force for the economy and society,
• A conversation with children and young people about AI through Comhairle na nÓg,
• UCC MOOC to deliver AI education to at least 1% of the Irish population,
• Funding to develop novel AI solutions,
• AI as part of future climate policies, with applications to monitor and reduce energy usage.

Challenges

“AI poses significant opportunities in addressing and overcoming pressing societal challenges, and creating new value and possibilities for everyone,” the minister says.

Ireland’s AI strategy must be human-centred, innovative, and governed to build trust and confidence that it can improve socio-economic performance and productivity, the document adds.

The strategy commits to an AI approach which is responsible, ethical and trustworthy

It says that much of Ireland’s existing legal framework should be sufficient to provide for the smooth deployment of AI.

However, some legal gaps may require additional regulation.

Bias

AI-based systems brings risks such as:

• Opaque decision-making,
• Discrimination,
• Bias,
• Privacy issues,
• Use for criminal purposes,
• Explainability, accountability and fairness.

The risk of discrimination arises from biased training data, biased design of algorithms, or biased use of AI systems, which could exacerbate existing structural inequities.

Public administrations worldwide are experimenting with algorithmic decision-making and predictive analytics in high-stakes areas such as policing, housing assistance, healthcare, and eligibility for social benefits, the document points out.

Algorithmic decision-making can also have significant effects in decisions concerning recruitment, or access to insurance and credit.

Therefore, information about how AI systems make consequential decisions must be made public and understandable, the document pledges.

Conformity assessment

‘Social scoring’ will be prohibited in Ireland’s AI regime. High-risk areas, such as recruitment, will also be subject to an ex-ante conformity assessment.

At EU level, development is underway of a horizontal European regulatory framework that protects fundamental rights, in an ‘ecosystem of trust’.

Mechanisms for voluntary and self-regulatory oversight will be considered for non-high-risk AI, such as pilot programmes, standards, labelling and certification schemes.

Other policy tools –  such as impact assessments, industry-led codes of practice, and ethical guidelines – will also be used.

'Smart mix' 

The document describes this ‘smart mix’ as a multi-faceted approach that will integrate laws, regulatory institutions and voluntary systems.

AI must support the ethical values of society and protect human rights, the document continues, and the Government will provide a range of tools to help in assessing the trustworthiness of AI systems.

Compliance tools, such as standards and certification, will be used to underpin both legal and ethical obligations.

This work will be led by the Top Team on Standards for AI, which was established in 2020 by the National Standards Authority of Ireland (NSAI) to develop a standards and assurance roadmap for AI.

Robust data governance

The document adds that robust data governance and a privacy framework are essential for a healthy AI infrastructure.

“Government has put in place legislation to allow public bodies to share personal data, which will be overseen by the Data Governance Board.

“Broadband roll-out and High Performance Computing (HPC) are long-standing Government priorities,” the document says.

The document commits to making more public-sector data available, and to facilitating trusted data-sharing mechanisms and research environments, to enable collaboration across different organisations.