- Hailey Halligan
Introductory
- Bank transfer
Event Organizer(s)
Description
As the volume of global data surpasses traditional analytical capacities, artificial intelligence has become an indispensable tool for modern policy makers. This course provides an introduction to the use of AI for policy. Specifically, it zooms in on how AI can assist by utilizing Machine Learning (ML) frameworks, from Natural Language Processing (NLP) and Large Language Models (LLMs) to Computer Vision (CV), to transform raw information into actionable policy evidence.
In this course, the focus is on
Nowcasting:
Modern policymaking requires timely and quality data. However, most policy-relevant indicators (from macroeconomic statistics to sector-specific information on employment, industrial activity, or innovation) are released with delays ranging from 30 days to several years. As a result, policymakers are often “driving while looking in the rearview mirror.”
The problem is even more severe for granular insights—such as regional dynamics, sectoral shifts, or the adoption of new technologies—where data may simply not exist.
AI-powered nowcasting addresses these gaps by leveraging proxy data and non-traditional sources to infer real-time conditions:
- Satellite Imagery: Nighttime light intensity to estimate economic activity; port and shipyard activity to track trade; land use patterns to monitor agriculture, infrastructure, and urbanization; housing characteristics to proxy inequality.
- Search Trends: Online search behavior (e.g., queries about unemployment benefits) to anticipate labor market dynamics ahead of official statistics.
- Textual sources: LLMs applied to news, central bank communications, patents, and job postings to extract signals on economic sentiment, technological change, and skill demand.
Evidence synthesis:
Policymakers face an overwhelming and rapidly growing body of evidence—from academic publications to policy reports and legislative documents. AI enables systematic and scalable synthesis of this information.
- Automated literature reviews using NLP and LLMs
- Classification and summarization of policy documents
- Extraction of key findings, consensus, and disagreement across studies
- Mapping of research landscapes and identification of gaps
These tools allow policymakers to move from anecdotal or selective evidence toward more comprehensive, transparent, and reproducible knowledge aggregation methodologies.
Causal Machine Learning:
While Machine Learning has traditionally been associated with prediction and often criticized as a “black box”, recent advances have increasingly integrated causal inference into ML frameworks. This part of the course focuses on how AI can move beyond prediction to support policy-relevant causal questions, such as: What is the effect of a policy? What would happen under alternative interventions?
Policy research examples
Our researchers inform participants not only on how their research uses AI to derive at new findings and conclusion, but also how this research is used to directly inform governments to influence their policy process
- Use of satellite data (Stephan Dietrich).
- Application of Natural Language analysis to innovation and employment (Tommaso Ciarli).
- Evidence Synthesis: Scan ODA landscape to understand how to make aid more impactful (Lindsey Moore).
Target population
This introductory course targets policymakers, government officials and other relevant civil servants from national and regional administrations, officials from international organisations, entrepreneurs, professionals in the digital and other relevant sectors, representatives of civil society organisations, universities, research centres, think-tanks, as well as the general public.
Kindly note this course is restricted to users who meet the following criterial:
Entry requirements
Members of the above-mentioned target population are invited to apply for the training if they meet the following criteria:
- Hold an undergraduate degree in a relevant field or have a minimum of three years of experience in the field.
- Possess a fluent level of English.
- Complete the application questionnaire.
- Government officials and policymakers from developing countries, particularly women, are encouraged to apply. We highly recommend the course to policymakers that have a research role or are tasked to use evidence in their policy making.
- Selection will be conducted by the course organizers, who will consider the above entry requirements along with an analysis of the application questionnaire of each applicant. The spot is secured upon payment.
Number of available places: 30
Learning objectives
Upon completion of this course, participants will be able to:
- Outline the various uses of AI for policy research
- Identify when a problem is suitable to de serviced with AI research, and when qualitative elements or human interpretation is key
- Compare uses and ways of evidence synthesis
- Use AI for nowcasting
Methodology
The course content includes (times to be confirmed):
- Unit 1 Lectures (synchronous): Tuesdays 8 September 13:00-14:00 CET and Thursday 10 September 13:00-14:00 CET
- Unit 2 Practical examples (synchronous): Monday 14 September 13:00-14:00 CET, Tuesday 15 September 13:00-14:00 and Wednesday 14 September 13:00-13:00 CET
- Required and optional readings for each unit
- Two Individual unit quizzes based on unit activities and materials. It is important for participants to read the session materials before the session starts, as the quiz opened up right after the lecture and will be proctured. Tuesdays 8 September 14:00-14:30 CET and Thursday 10 September 14:00-14:30
- Individual reflection assignment
- Final pitch of the individual assignment
All materials can be accessed through the ITU Academy platform.
Assessment and grading
Participants are required to submit one multiple-choice exam at the end of each module. Each assignment is graded individually on a 1-10 scale.
Upon completion of the course, participants need to submit an individual policy memo.
The final grade for this module is the simple weighted average of all graded components’ grades
- Unit 1 Quiz : 20%
- Unit 2 Quiz : 20%
- Unit 3 Individual Reflection Assignment on how to use AI for research within your own policy environment : 60%
- Unit 3 Individual Assignment Pitch : Pass/Fail
You will pass the course if the total weighed average grade is 55% or higher and if you pitched your reflection assignment to your peers.
Participants are required to attend both lectures of Unit 1 at least two of the three case sessions from Unit 2. and attend the session in Unit 3. Participants can attempt each Unit Quiz only once, right after the lecture. For the individual assignment, there is one resit opportunity. The pitch to the peers is pass/fail and mandatory. No resit is offered for the pitch.
You are allowed to take the resit if your final assignment grade is below 70% AND if your final grade is below 55%. The total successful completion score for certification needs to be a minimum of 55% or higher (on a scale of 0-100%).
Kindly note that this course is introductory and we aim to inform you on the potential use of AI for policy research. We do not expect detailed proposals on how to use AI in your context, rather an informed thought-memo on what could be interesting applications. The course is specifically tailored to fit the needs of people that have not yet worked with AI.
Training details and instructional approach
The course is divided into units. Unit 1 includes the theoretical sessions, which are complemented with practical examples in Unit 2.
The units lectures are:
Unit 1 lecture 1: Fundamentals of ML for policy making, nowcasting, forecasting, causal inference.
Unit 1 lecture 2: Fundamentals of ML for policy making, nowcasting, forecasting, causal inference.
Unit 2 lecture 1: From unstructured to structured data: Computer Vision.
Unit 2 lecture 2: From unstructured to structured data: Natural Language Process
Unit 2 lecture 3: Evidence Synthesis
Unit 3 sessions: Pitch of individual assignments
Registration information
Unless specified otherwise, all ITU Academy training courses are open to all interested professionals, irrespective of their race, ethnicity, age, gender, religion, economic status and other diverse backgrounds. We strongly encourage registrations from female participants, and participants from developing countries. This includes least developed countries, small island developing states and landlocked developing countries.
