The importance of capacity building in the development and use of AI systems
Capacity building programs should be put in place to ensure that stakeholders, particularly those in developing countries, have the knowledge and skills necessary to engage with AI and understand its potential impacts.
Generally, it is accepted that AI deployment in developing countries will create new employment opportunities, hasten organizational efficiency, and enhance public service delivery. However, skills gaps and infrastructure impede the capacity of developing countries to tap into AI’s potential.
In an already unequal society where employment opportunities are often less accessible to people in resource-constrained settings, the efficiency-related advantages of embedding AI into existing systems will also disregard these populations. In addition, as it is happening in other world parts, AI introduction is likely to result in crucial challenges, including misuse or uncritical acceptance of biased automated decision-making.
The capacity to face these challenges will mainly be impacted by the level to which countries have nurtured a skilled workforce with technical skills, training, and knowledge to govern AI deployment.
Thus, efforts should be made to build the capacity of individuals and organizations working in AI development and deployment. This can involve providing resources and training to help them build their skills and knowledge in AI.
For example, knowledge scientists should be trained to enhance the decisions driven by the AI system. In addition, as AI automates mundane and repetitive tasks that are done by humans, it is critical to train individuals on how to interact with these intelligent systems.
AI literacy is a major element required to upskill workers and managers interacting with AI systems. It necessitates knowledge workers to have a better appreciation of their artificial counterparts, algorithmic competencies as well as new analytical, data-centered skills that enable workers to interpret artificial intelligence-based decisions.
Thus, stakeholders should have a curious mindset for asking questions, critically and actively engaging with algorithmic results, and providing feedback that can be deployed for the AI system. These can be achieved through the capacity building of individuals and organizations who work in developing and deploying AI.
Recommendations for ensuring capacity building in the development and use of AI systems
Improve Data Availability: Training on the importance of open data should be done, particularly for public sector employees. Concomitantly, training on the importance of transparency is required to instill knowledge on how AI is being deployed and the measures for mitigating its harms and risks.
Socioeconomic Risk Assessments: They should be performed prior to AI development. In situations where decisions have already been made to spur the use of AI, they should be preceded by socioeconomic risk assessments to enhance awareness of the implications.In situations where decisions have already been made to promote the use of AI, socioeconomic risk assessments should be conducted first to raise awareness of the potential consequences.
AI Governance Skills: Capacity-building programs for AI often focus on technical skills with little attention to governance issues. However, to ensure that AI is implemented or developed with due consideration of the potential harms and risks it poses, it is important for capacity-building organizations to embed AI cybersecurity training, ethics, and governance as critical elements of their curriculum.
Reduce Socioeconomic Inequalities: Public resources should be invested to reduce social and economic inequalities. Tackling these inequalities would play a critical role in reducing the probability of AI-powered systems aggravating the existing inequalities. For instance, developing countries should invest in enhancing the quality of education for all to lay the basis for embedding AI into the economy more seamlessly. It would be crucial to ensure that existing imbalances in AI-specific skills do not continue being barriers/impediments for disadvantaged populations to join the AI workforce.