"In 2024, we expect the ongoing digitalization process to further improve program efficiency and the effectiveness of public policies in Portugal. Two critical elements driving this digital transformation are data and artificial intelligence (AI). AI can play a central role in creating value from data and extracting deeper insights from the vast information public administrations gather to serve their citizens.

As demand for generative AI is expected to grow this year in Portugal and worldwide, it becomes imperative that public sector leaders also embrace the responsible use of this technology. Only then can they establish themselves as trusted guardians of information.

The differences between generative AI and traditional AI

To understand the unique challenges posed by generative AI in comparison to traditional AI, it is important to understand their fundamental differences. Traditional AI relies primarily on algorithms and large sets of labeled data to train models through machine learning. These models can make recommendations or identify certain behaviors by recognizing patterns and following predefined rules. For example, traditional AI is used to improve the effectiveness of email spam filtering, enhance movie or product recommendations for consumers, and enable virtual assistants to help people find information.

Generative AI is emerging as a valuable solution to automate and enhance routine and repetitive administrative tasks. This technology differentiates itself by leveraging foundation models, which are large neural networks trained on vast datasets of unlabeled data and fine-tuned for various tasks. In effect, it can effectively identify, summarize, translate, predict, and generate content from large datasets. Implementing this technology in the public sector can significantly improve efficiency, enabling organizations to complete their daily tasks with a fraction of the resources.

Generative AI thus presents an unprecedented opportunity to improve various aspects of public sector operations and services for citizens. It can empower public servants with more powerful tools to answer questions and conduct research. Other tasks such as drafting and managing contracts, which are both time-consuming and crucial, can also benefit from the application of generative AI.

Gonçalo Costa Andrade© IBM  

Implementing generative AI responsibly

The remarkable generative capabilities of this emerging AI technology, however, raise questions about its responsible use in the public sector as well. For example, public contract managers need to know that the research that went into it has been faithfully translated into a legally binding contract for two or more parties.

People have only recently begun using generative AI, primarily through tools that use preexisting text, images, videos, and audio to create custom content based on a given prompt. However, the level of detail that goes into training some of these models may not be sufficient, especially for large enterprises or highly regulated industries that depend on public trust.

To develop responsible AI, public sector leaders must carefully curate their data to get the most value from both AI and generative AI. Setting responsible standards is a critical role for public administration, and it requires building that responsibility in from the beginning, not as an afterthought. This includes maintaining human oversight to help ensure the accuracy of AI-generated content and avoid bias, among other issues.

Key pillars for responsible AI in public administration

IBM’s development of AI is centered around five key pillars that help ensure trustworthy AI. Public sector leaders should prioritize these pillars whenever they consider the development, training, and responsible implementation of AI:

1. Fairness in an AI system refers to its ability to treat individuals or groups equitably, depending on the context in which the AI system is used. This means addressing bias and preventing discrimination that relates to protected characteristics, such as gender, race, and age.
2. Privacy concerns an AI system’s ability to prioritize and safeguard consumer data privacy and rights, while complying with existing regulations related to data collection, storage, access, and disclosure.
3. Explainability is important because an AI system should be able to provide a human-interpretable explanation for its predictions and insights, in a way that is not obscured by technical jargon.
4. Transparency means that an AI system should include and share information about how it was designed and developed, as well as the data or data sources used to train the system.
5. Robustness is the ability of an AI system to handle exceptional conditions effectively, such as anomalies in data input. It helps ensure consistent results.

IBM WatsonX™ is an integrated AI, data, and governance platform that incorporates these principles, providing an integrated, efficient, and responsible approach to AI development across multiple environments. More specifically, the recent launch of IBM® WatsonX.Governance™ helps public sector teams automate and address these areas, enabling them to guide, manage, and monitor their organizations’ AI activities. This tool facilitates clear processes so that organizations can proactively detect and mitigate risks while supporting their compliance programs for internal AI policies and industry standards.

As public sector leaders in Portugal continue to seek ways to deepen their adoption of AI and automation to solve problems and improve efficiency, it is essential to maintain trust and transparency in any AI solution. Teams need to have the ability to understand and manage the AI lifecycle effectively. Proactively adopting responsible AI practices is a unique opportunity for all of us to do better."

Read Also: AI: Rule of Law may collapse if there are no limits to technologies (Portuguese version)