Jordan’s AI policy journey: Bridging vision and implementation

By Annys Rogerson

Jordan is at a turning point. The Jordanian Government established a dedicated AI division within its Ministry of Digital Economy and Entrepreneurship (MoDEE) in 2020, headed by Lama Arabiat (Eng). Since then, the division has been focused on developing its National AI Strategy and AI Ethics Charter, both of which were published in 2022. Now, the division is turning towards the implementation of its plans and is looking towards the challenges involved in doing so.

Developing Jordan’s vision for AI

Jordan took a collaborative approach to the development of both its national strategy and ethics principles, involving representatives from across government, the private sector, academia and NGOs. Jordan took this approach to ensure that both documents were tailored to the needs of Jordanian society, public sector, and businesses, and so that its plans were achievable for the country’s AI community. The collaboration culminated in the national strategy that lays out a roadmap for the next 5 years. The strategy includes 5 main strategic objectives:

• Capacity building, and developing Jordanian skills and expertise in the field of AI
• Encouraging scientific research and development in the field of AI
• Enhancing the investment and entrepreneurship environment in the field of AI
• Ensuring a legislative and regulatory environment that supports the safe use of AI
• Apply AI tools to increase the efficiency of the public sector and priority sectors

Within the roadmap, the government commits to an ambitious 68 projects, across 12 priority sectors. These projects will be coordinated by MoDEE’s AI division and delivered through collaboration between public, private, academic, and NGO partners.

Notably, only 35% of countries who have a National AI strategy have also adopted a set of national AI Ethics Principles. Within Jordan’s income group, the proportion of countries with ethics principles drops to 10%. Jordan chose to develop a National AI Ethics Charter concurrently with its strategy so that its burgeoning AI community would have a set of basic principles to form around. Given that ethics principles at a country-level are more common within country’s who are further along in AI adoption, Jordan’s early development of principles serves as a trial for the impact of ethics principles on nascent AI communities. The outcomes of this trial will depend on the adoption of the ethics principles by the community, which is now a focus for the MoDEE’s AI division as it begins implementing its vision.

Implementing Jordan’s vision for AI

National AI Strategy

As a first step towards implementing Jordan’s National AI Strategy, the MoDEE carried out AI readiness assessments on its governmental institutions. The assessments were done on 18 governmental institutions and came up with a gap analysis and recommendations for each entity in addition to an institutional 5-year action plan for adopting AI. The analysis found that the MoDEE was the most ‘AI ready’ agency and data quality was identified as an area for improvement across government. A secondary aim for this project was to raise awareness of AI among public sector workers. Lama Arabiat said that assessors recorded a 25% increase in awareness of AI among staff and over 3000 staff members attended the capacity building AI workshops involved. She suggested that the buy-in for AI adoption among agencies as a result of the assessments will be a key enabler for them in the years to come.

Promisingly, Jordan is now taking a cross-government approach to improving data quality, with MoDEE being responsible for creating a central, ‘national information system’ that collects data from several government entities. This is a crucial initiative for the adoption of AI in the public sector because, in many cases, AI systems will require data from a range of entities that needs to be structured according to the same standards. As well as putting standards in place for data quality, these efforts should be accompanied by strong data governance practices so that citizen’s data is not shared or used without explanation, consent, and mechanisms for redress.

The AI readiness assessments have also helped the government with the next stages of implementation, delivering their 68 projects, by identifying opportunities for adopting AI in the public sector. Jordan has started a pilot project to deploy AI to reduce traffic congestion. This project is led by MoDEE in collaboration with public, private, and academia partners. Currently, the team is at the early stages of piloting the project in a small geographic area in order to create a proof of concept they can scale up.

National AI Ethics Charter

Implementing principles of ethical AI within a country is less well-trodden ground. There is a danger that a government creates commendable principles of ethical AI but there is limited adoption of those principles within the country’s AI community. Jordan is now facing the task of avoiding this danger.

Jordan’s Ethical AI Charter is intended as a guideline for developing and using AI in Jordan and it is not binding on any groups. For now, the government’s approach is to encourage the adoption of its ethical principles among the AI community through:

• Raising awareness through school and university curricula as well as industry seminars and workshops,
• Exploring the creation of awards, accreditations, and incentives, such as tax credits, to encourage companies’ compliance to the principles using a specialised body capable of evaluating performance, and
• Implementing policies that support the adoption of the principles, such as creating an open government data initiative that provides local researchers with access to more data and data that meets standards with respect to representativeness.

Importantly, the Charter fits into the government’s broader efforts to support responsible use of digital technologies in Jordan. This work includes developing cybersecurity and data protection and privacy legislation, which will be binding. Lama suggests AI-specific legislation could be adopted later down the line and that the principles are a useful step towards understanding what this legislation would look like. For example, reports of violations of the principles could point to where legislative constraints are needed.

Ensuring progress

As Jordan embarks on its 5-year plan for supporting AI development, monitoring and evaluating progress will become crucial. The responsibility for monitoring progress, Lama Arabiat comments, will be with MoDEE and the AI division. In order to maintain citizens’ and industry’s confidence in the government, it is important that this monitoring is done openly. Openness is needed for citizens to continue to trust the government’s use of AI and industry professionals require the certainty that investments are being made as promised. Efforts to encourage openness include publishing progress on the strategy’s key metrics and adopting mechanisms that support transparency of AI use in the public sector.

More generally, when we monitor progress of government AI projects, we should recognise that using an AI system may not end up as the best solution to a problem. Governments can hugely improve public services through using data more effectively without the use of AI techniques and this should not be considered a failing. Therefore, project teams should remain open to solutions that do not involve AI. This should be recognised when Jordan monitors the progress of their own projects.

  Finally, as Jordan progresses with its projects, the government will itself be grappling with the challenge of putting its ethics principles into practice. This is an opportunity for the government to support the adoption of the principles outside of government too. Understanding best practice from its own experience will help the government to communicate to industry how to comply with its Ethics Charter. Consequently, the government can share its learnings through its awareness raising initiatives and feed its learnings into how it evaluates compliance within the AI community.

___

Should we be interviewing you next? Contact us at  research@oxfordinsights.com