Interview Guide

Interview Guide, EMB Country Profiles

The following document details the interview structure and questions for the country case profiles drafted for the Microsoft-hosted AI & Elections Hub. For a thematic overview of the profiles’ content and structure, please see the Case Framework document.

While the interviews will be predominantly based on the questions featured below, their semi-structured nature will allow for follow-up questions that may ask the subject to further expand, clarify or add to any points. There may also be additional questions on the contextual electoral experiences of the specific country which an interview subject is representing, i.e. a Swedish representative may be asked questions relevant to the Swedish electoral system or experiences in the country’s recent electoral cycles.

Case Framework Document Profile Framework

Interview Questions

The interview questions are divided into thematic subheadings that follow the structural outline of the profile framework. The interview process will follow this structure chronologically and will omit any non-applicable or irrelevant questions from the interviews on a case-by-case basis. Aside from gathering some introductory background information, the initial questions serve to establish a baseline rapport with the subject before broaching the subject of AI. The duration of each interview will span approximately 60-90 minutes.

The interview process will follow this structure chronologically. The duration of each interview will span approximately 60-90 minutes.

1. National Electoral Environment

Can you briefly describe the institutional mandate of your organization?
How would you describe the level of digitalization relevant to elections in your country? Could you highlight any particularly developed areas or any gaps? Including both wider digital public services (i.e. digital identification services, public agency communications, etc.) and within election administration (i.e. voter databases, institutional cybersecurity capacity, e-voting systems, etc.).

How accessible are these services to citizens in practice, especially outside of major cities?
What is the decision-making structure for introducing new electoral technologies? Who has authority in approval processes?
What recent events in your elections are most relevant for understanding your organization’s current approach to AI?
Have there been any major incidents, such as other actors using AI in the electoral sphere or the introduction of new regulatory protocols, that have influenced your approach?

2. Use of AI to improve Electoral Administration

What AI-enabled tools or systems are your EMB currently using, piloting, or actively exploring? If you have yet to explore any concrete options, how have you discussed the possibility of implementing AI for your work?

If yes, follow up:

Can you walk us through a concrete example where your EMB identified a problem or process and decided to use an AI tool to address it? From initial need to implementation and evaluation.
Are these systems developed in-house, collaboratively with other institutions, procured from vendors, or adapted from tools used by other public sector bodies? If they are externally developed, what is the institutional nature of the main provider (e.g. private company, NGO, semi-public entity, international partner)?
Do you have any internal procedures to assess the benefits and possible risks of using AI instead of traditional approaches for different electoral applications?
Have you encountered any positive or negative unexpected consequences in deploying AI? For example, any systematic error or bias, changes in capacity or skill needs, public trust concerns, or security and/or privacy risks?

If no, follow up:

Is your EMB considering implementing the use of AI in the future?
What would you say are the main obstacles, concerns, or challenges that preclude your EMB from implementing AI tools? What are some key institutional prerequisites that would precede the introduction of AI tools?
Can you identify any electoral management processes that may be improved by the introduction of AI tools?

What types of resources (knowledge, capacity, guidelines) do you believe are most crucial for your EMB to develop your approach to AI? In what way do such resources support your work on AI?

Looking ahead, what area of your work do you see as most likely to involve AI systems in the future? Which do you think are least likely (or appropriate)?

3. Institutional Adaption for Trustworthy AI

Has your organization undergone any institutional changes to support the adoption of AI? I.e. any changes in your mandate or your internal operations and structure?

If no: Do you have any plans to do so in the future? Why or why not?

How has your institution responded to the use of AI by other electoral actors (including campaigns, independent political influencers, media outlets)?

Have you had to adapt to any new or existing national regulatory instruments that apply to your work relating to AI? Have you introduced (or are you considering) new internal guidelines, standards, or evaluation procedures specific to AI use?

If yes: How has your EMB engaged with the development of electoral AI regulation on a national level (incl. drafting, consulting, and developing)?

Do you have (or are you considering) any dedicated, unit, task force, working group, or similar coordination mechanism that manages AI-issues? If so, how do they function?

Do you have any specific initiatives to train staff in understanding, managing, or overseeing AI systems? How are you discussing AI literacy and responsibility with internal staff?

4. Safeguards for Ethical AI Adoption

What ethical, regulatory, or human rights considerations would lead your EMB to decide not to adopt AI for certain applications?

How does (or would) your EMB assess the ethical and legal risks before the deployment of any AI system?

How does (or would) your EMB evaluate and address the risks during and after the deployment of any system? How would you ensure human oversight and compliance with ethical standards?

How have you considered any possible risks of discriminatory or exclusionary bias that might arise from the use of AI in certain processes? Do affected constituents have any means to seek redress if they are negatively affected? What would redress mechanisms look like?

How do you ensure that AI use is transparently communicated to the public in an accessible and understandable way? Do you offer alternative access to services that involve AI tools for constituents that wish to opt-out of its use?

What is your approach to voter data collection, management, and privacy? How do you ensure the security of sensitive voter information?

5. Takeaways

If you were to advise electoral management bodies in other countries in your region who are approaching AI in elections, what central lesson based on your experiences would you like to share?