Election Readiness - Gotowi na wybory! - supporting people who run and observe elections
Share project
Table of contents
- The problem
- What's already in place
- Project scope
- Users and functionalities
- Data
- Desired expertise
- Success vision
- Supporting resources
The problem
Problem statement
We, the Political Accountability Foundation (in Polish: Fundacja Odpowiedzialna Polityka i.e. FOP) create training materials for election commission members and election observers. Every election cycle we struggle with updating the materials according to new guidelines and code on short timelines. Our solution helps us automate the process and reach the learners in a way that they use the materials.
The timing Training for Poland's polling-station commission members and election observers must stay 100% faithful to the National Electoral Commission's (PKW) guidelines and the Electoral Code — sources that typically change only 3–4 weeks before voting day, i.e. exactly when the Foundation's workload peaks. Guidelines can also be challenged before the Supreme Court after training has already started, leaving some commission members to update their knowledge entirely on their own — not all of them manage to.
The format Today that update is manual, slow and error-prone: there is no repeatable process that shows, for each piece of content, which provision it comes from, which version, and who approved it. A second, compounding problem is format: existing FOP "checklists" run 8–13 pages at 30–37 words per sentence, while mobile traffic rose from 31% (2023) to 57% (2025) and the entire value of the material is used inside a roughly 72-hour window around voting day.
Recurrent nature without an appropriate official solution The official training is not aligned to learners' needs. The training deck has been essentially unchanged for 11 years (176 slides). This is a recurring, structural bottleneck, not a one-off — it repeats every election cycle, and directly limits how many first-time and last-minute commission members and observers FOP can reach with accurate, usable material in time.
What's already in place
Current tech stack
- Web application
- Frontend or user interface
Impact is that we have a way to create materials that need updating on short timelines and a way to show it on a mobile to a learner.
Internal engine A reviewer (FOP content owner) selects an audience (commission member or observer), a stage of voting day (A–H, 8 stages) and a material type (checklist, quiz question, interactive scenario). The engine proposes content strictly from three approved inputs — training objectives, FOP's own materials, and the legal sources (National Election Commission regulations and the Electoral Code). The reviewer approves, requests a correction, or rejects each item individually (corrections and rejections require a justification). At this stage aim to go through series of such a cycles so our agent will be tailored with language, tone and ways of creating content that FOP team fully accepts. Only approved items are published.
Learner friendly material (external facing content) For the end learner, today a microlearning unit takes 10–12 minutes and combines one scenario, five questions and a checklist-ordering exercise (keyboard-accessible).
MVP still has tags: every item carries a double tag: content provenance (FOP course material vs. engine-generated vs. approved-bank item) and review status, so nothing unreviewed can be mistaken for finished material. Final product for end learner shouldn’t have it.
Project scope
Workstream 1: product hardening and cloud deployment
- Move the prototype from its current free-tier setup to a production-grade cloud architecture that scales up for a 72-hour election peak and scales down afterwards.
- Secure multi-user login, user management and a managed database in place of browser-held state.
- An administrative console for content generation.
- A review console for validating generated questions, quizzes, checklists and scenarios, supporting several reviewers working in parallel with persistent work in progress.
- Monitoring and analytics sufficient to understand usage and system performance.
Workstream 2: AI-assisted regulatory content audit
- Compare the expert-verified corpus against new versions of the NEC guidelines (195 points) and the -Electoral Code (611 fragments) at the level of the legal provision.
- Identify which materials are affected, flag the relevant differences and suggest where content should be updated, removed or added.
- Produce a "what has changed" summary between versions for learners as well as reviewers.
- Use code to update the existing content when changes are required
Workstream 3: mobile-first microlearning site for learners
- A learner-facing site designed from the phone screen outward, since 57% of our traffic is now mobile and it arrives inside a 72-hour window.
- Microlearning units of 10 to 12 minutes: a scenario, five questions, and a drag-and-drop ordering of a checklist, with keyboard operation for users who do not use a mouse.
- A one-page checklist card as a distinct material type, usable at a polling station at 5:30 in the morning.
- Accessibility built in from the start. Every interactive package we audited lacks heading structure and a declared document language and permits zoom to be disabled on mobile. The new layer must not inherit these defects.
- Learning analytics: what people open, where they stop, which questions they fail. This is how we improve materials with evidence.
Users and functionalities
From creation side (FOP experts - login required): FOP content reviewer (subject-matter expert): selects audience / day-stage / material type, reviews AI-proposed items one by one (approve, request correction, or reject with justification), and publishes approved items with version, date and owner attached.
Clients (wide audience - publicly available - advice needed on data collection available with and without login):
Persona 1 - First-time polling-station commission member: works through microlearning units matched to his role and the day's stage (scenario + 5 questions + checklist-ordering, 10–12 min).
Persona 2 - election observer: same learning mechanics, scoped to the narrower, boundary-aware observer role (witness, not judge).
Data
Data Readiness
- Spreadsheets or documents
- Third-party systems
Desired expertise
Tech partner role
- Frontend / full-stack developer comfortable with a lightweight, dependency-light stack and with accessibility (WCAG) for a mobile-first, no-login product.
- Someone who can add minimal persistent storage for review-session state without introducing personal-data handling.
- Data/analytics support to instrument review-time measurement and reach metrics.
- UX/product designer for the phone-first one-page checklist and for accessibility remediation.
Success vision
Success means we protect our resources during the crunch time election window to look at updated content only, all while rapidly delivering a learner experience that matches learners needs (format, availability) and is expert approved.
-
Training content updates on a predictable, auditable cycle Our agent tailored to our needs in terms of language, tone and ways of creating content that FOP team fully accepts generates materials so its acceptance is just a fast track pro-forma instead of long dispute and comments. It means that previously developed content does not need to be changed unless changes in Code/Guidelines warrant it.
-
Training available at on a mobile to large numbers of learners during the election peak In outcome terms: commission members and observers are able to control by themselves learning process and check if they are ready for the role (today there is no test item or check point for those who need it. Moreover those who join late or whose training happened before a Supreme Court correction can catch up on their own instead of going in blind; disputed on-the-day situations are handled from rehearsed scripts instead of improvised passivity or overreach.
Supporting resources
.png&w=3840&q=75)
