The issue isn’t always whether data is public. Often it’s right there on a government website, yet still inaccessible, confusing, or impossible to use. That’s why Political Watch built “What Our Representatives Do” (Qué Hacen Los Diputados, or QHLD), a platform designed to make the work of the Spanish Congress understandable and searchable. 

During Spain’s housing crisis—with young people unable to leave their parents’ homes, families crowding into shared apartments, and evictions occurring daily—QHLD helped reveal a striking mismatch between political attention and public need. We found 198 legislative initiatives on squatting, an issue affecting just 0.06% of the housing stock, but only 54 initiatives on social housing, the area at the heart of the crisis.

This imbalance exposed an apparent disconnect between the political establishment and the people filling the streets, waving their keys, and demanding fair housing policies. 

QHLD made this discrepancy visible. Maintained by Political Watch since 2011, the tool organizes and presents congressional activity in an accessible format, enabling searches by topic, date, author, and cross-reference across 24 social policy areas spanning the environment, economy, and democracy. It remains the only tool of its kind in Spain.

And it does more than track legislation; it also monitors parliamentary initiatives to strengthen oversight and accountability.

Focus on Usability

The QHLD’s strength isn't providing new information, but organizing what already exists in ways that are comparable, downloadable, measurable, and reusable. We add value to data that already exists, but doesn't effectively illuminate parliamentary dynamics.

From the outset, its goal has been to defend the role of parliaments amid increasingly presidentialist leadership. Although Spain's political system isn't conceptually presidentialist like the United States, it has recently trended toward executive decree-based governance with minimal honest debate, copying the worst aspects of presidential systems. It follows the same approach as TheyWorkForYou, designed by mySociety for the UK, or the newer Mzalendo, which tracks electoral promises in Kenya's Parliament.

We want to bring voters closer to what their representatives actually do, attempting to bridge the growing gap between representatives and the represented—a goal we also advance through strategic alliances with other organizations and collaborative advocacy. 

This year, we helped secure approval of the first Open Parliament Plan in the history of Spanish democracy.

But QHLD is not an end in itself. At Political Watch, we monitor power as a step toward sharing it. We advocate for more participatory governance, where citizens are involved in decision-making on issues that affect them—both indirectly through representation and directly through mechanisms such as assemblies, participation platforms, petitions, and public consultations. We defend the democracy we have while building the democracy we need.

Key Features

Among QHLD's main features is thematic tagging, designed with a co-creation spirit alongside civil society. In fact, the more than 5,000 concepts the platform uses have been developed over the years in collaboration with over 50 social organizations and subject-matter experts. This system allows initiatives to be linked to multiple themes, fostering cross-cutting analysis.

Additionally, we've developed a ranking that scores each representative based on weighted criteria: their volume of parliamentary activity, their ability to build consensus to pass initiatives, their level of action on each issue (inaction is penalized), and how easily citizens can contact them.

A distinctive feature is the ability to receive personalized alerts that users can activate based on their interests (user-centered, not spam). Another key element is that the system is 100% developed under free and open-source software principles. This has enabled its implementation in other countries, such as Paraguay, in a short timeframe and at very low cost.

Use of AI: Still a Long Way to Go

While powerful, the tool has also faced challenges in scaling its use.

Right now, Qué Hacen Los Diputados relies on classic natural language processing (NLP) to determine how closely each parliamentary initiative aligns with specific topics, generating a percentage match. It works efficiently, especially given the dataset's size.

We’ve experimented with incorporating generative AI, but haven’t fully integrated it yet for two main reasons:

1. Cost barriers for this specific use case.

Generative AI is expensive, not because of the tool itself, but because Spanish parliamentary work produces thousands of initiatives and large volumes of documentation, often lengthy PDFs with very high token counts. As a nonprofit project with no commercial model or dedicated funding, adopting a high-usage, high-cost system is risky, particularly at the scale required for QHLD.

2. Limited added value for our current workflow.

Our system classifies initiatives into 24 thematic categories, a relatively large number for any AI model. Our current NLP method (similar to a bag-of-words approach) performs reliably enough for our needs. While this approach limits semantic nuance, we’re exploring ways to add semantic depth through embeddings and related techniques.

Where generative AI will play a major role is in the next significant upgrade to the platform, a new semantic video search engine currently in development. We are excited to experiment with this tool and document the lessons as we go. 

Dilemmas

Finally, our ethos is to share our progress without hiding the dilemmas we face. For example, why can't we drive more traffic to the website, even though it's free and doesn't even require registration? So far, we know that around 500 NGOs, academics, and journalists use it regularly. Still, we haven't managed to spark interest among broader audiences, not even among organized groups like journalist associations.

Should we implement access restrictions, now that we've detected commercial use by large public affairs consulting firms? Should we dedicate more time to fundraising to enable innovation? Should we abandon the use of classic NLP technology, which gives us such high reliability in results, for more advanced and efficient approaches that we're unsure will perform as well?

The list of improvements Political Watch wants to introduce to QHLD is long: semantic search of representatives' video interventions, parliamentary spam identifier, a calculator to determine the probability of bill passage based on historical party agreements on specific topics, a chatbot that delivers answers and solutions rather than just results, and much more. 

 However, resources for these developments are scarce, so innovation is, unfortunately, very slow.

Civic technology knows no borders. The challenges facing "What Our Representatives Do"—financial sustainability, mass adoption, the balance between openness and commercial use—are the same as those faced by similar projects in democracies worldwide. Perhaps the next innovation in parliamentary transparency won't require a large budget, but rather collaboration among small teams that, from anywhere in the world, share code, dilemmas, and solutions. 

Ultimately, monitoring power isn't a national project, it's a global democratic imperative