One thing we want to take more advantage of with TheyWorkForYou is the fact that we’re not an official website — and so can pull on multiple official and unofficial sources of information to present a richer picture of how our democracy works. 

Our trajectory with voting summaries has been to focus on votes that are substantive. This means they’re generally on issues whipped by parties, and there are few differences between the voting records of MPs in the same party.

But we’d also like to make it easier for everyone to understand what differentiates MPs: the signals they give about their values and interests, and where they fall on internal arguments about policy direction. 

As such, all MPs now have a Signatures tab on their TheyWorkForYou page, which tracks Early Day Motions (EDMs), open letters, and Motions to Annul signed by the MP. 

EDMs

One form of information we want to make more use of are Early Day Motions (EDMs). These are technically ‘proposed motions’ that may be elevated to a full debate. In practice this rarely happens and they work as an internal parliamentary petition service, where MPs can propose motions and co-sign ones proposed by others. They are still useful in reflecting the interests of different MPs even if EDMs rarely lead to substantive change in themselves. 

To provide better access to this information, we’ve added EDMs to TheyWorkForYou Votes as ‘Signatures’. Here TheyWorkForYou Votes is working as a general data backend that will help power features in our own services, and makes it easier to access the data for bulk analysis. This then feeds into individual MP profiles. 

With this, we are catching up to what Parliament displays on their MP profiles (EDMs), but also building the framework to expand to the UK’s other Parliaments and to capture extra-parliamentary statements like open letters that serve a similar function. 

Open letters

Over the last few years, we’ve noticed more open letters being shared on social media, where screenshots of a list of names on official parliamentary paper are serving the purpose of  signalling in public that a grouping exists in a political argument. 

A recent example of that is the big open letter for UK recognition of a Palestinian State. This was initially posted on X as images, and we’ve transcribed it and made the list of MPs searchable. 

There are a few reasons why MPs might prefer to use these kinds of open letters rather than submitting an EDM. Social media reach means that MPs can make a full public statement without the parliamentary publishing process. A letter can be published in full without the word count restriction of a letter to a newspaper, so can pick up more names.

Similarly, open letters are free from the format restrictions and word count of EDMs (a single sentence of less than 250 words). This can be important as many letters represent a group of government MPs trying to change the government position. Being able to write more is important in referencing previous government actions, anchoring the change in agreed principles and so on,  while still being a critical signal. 

This fits with a general change in usage of EDMs. While the number of actual EDMs proposed per year  have remained roughly the same, overall signatures have dropped by almost half since 2015 (33k to 15k), and far fewer petitions get a large number of signatures. The average number of signatures per EDM has dropped from 27 to 12. Some of this activity has moved to the new social open letter format. 

There are also some disadvantages to open letters. Publishing via screenshots means it’s not very accessible or searchable — a problem if one reason for signing is to signal to constituents.  If an open letter is important, people want to sign after the fact. EDMs have a mechanism for that, while for open letters you might get “here’s another page of names in another tweet” or social media posts saying “I support this too” —  but not in the same place as the original. 

For our purposes, it also means there’s collection work to be done finding the letters in the first place, and transcribing the images into text. We’ve got some good technical processes on the latter; and we’ve opened a form here where people can tell us about them. But it’s more work than just plugging into Parliament’s feed, which is what we do for data elsewhere on TheyWorkForYou. 

Looking at open letters is a shift towards including more extra-parliamentary activity — but reflects the need for parliamentary monitoring sites to react to changes in how parliaments and representatives behave, and think creatively about how to make use of new sources of information. 

Motions to Annul

Motions to annul are technically a form of EDM, but we’ve separated them out because we see them as something worth highlighting in their own right. 

To take a few steps back, when Parliament passes laws (primary legislation), it fairly commonly gives the government authority to make additional orders/regulations (secondary legislation) that fill in specific details in laws without the full parliamentary process. 

Secondary legislation still needs to be approved by Parliament – and this happens in two ways depending on how the law was written. Either the regulations need to be approved in a vote to become law (positive procedure), or they need to not be voted against within 40 days (negative procedure). 

Most legislation (around 75%) is passed through the negative process, and in practice the power to object is used very rarely (the last successful Commons objection was in 1979).

The mechanism is to make a Motion to Annul (for historical reasons called a ‘prayer’) through the EDM process. There is no threshold at which this is promoted to a vote and the government controls the Commons agenda. It is more likely if the motion is tabled by the Leader of the Opposition, or as the number of signatures goes up.

Come to our event

Join us on Thursday 23 October for a webinar on our new features, plans for the site, and our vision of a more open Parliament. 

Even if rarely successful,  these represent engagement with the legislative scrutiny process, which we felt was worth highlighting, and we separate these out in the signatures page from other EDMs. 

If you’ve ever wondered what your MP is interested in outside of their party alignment, a good place to look is All-Party Parliamentary Groups (APPGs). These groups bring MPs and Peers from different parties together around shared policy interests. 

There’s a real range of causes, of size, and of activity. The groups don’t have formal powers, but they can be influential spaces for discussion and collaboration. This also makes them key sites for lobbying, and for money to enter Parliament. For all of these reasons, as part of our WhoFundsThem work looking into MPs’ financial interests, we’ve been digging deeper into APPGs. 

Alongside our regular output that makes it easier to compare each APPG register to the previous one, there’s now a big new update on TheyWorkForYou allowing you to browse your MP’s memberships for the first time.

Getting the lists

There is no central list of memberships of APPGs. The official Parliament register lists the four officers of each group, but not the wider membership list (each group must have at least 20 members to be constituted). Some APPGs have websites where they publish these lists, but others don’t have public membership lists at all.

Two things have changed in the last few years that made it practically possible to put together a (mostly) comprehensive membership list. 

The big one is that the rules changed so that APPGs need to either publish a membership list on their website or provide it on request. 

The second is that LLM technologies have made more flexible scrapers viable, meaning we can more easily extract membership lists published in lots of different forms on lots of different websites. 

We’ll write up the scraper in a technical blog post,  but by scraping the available websites and requesting the membership lists from the remaining groups, we’ve brought all of this information into one place. 

Theory vs practice

From our previous experiment asking APPGs for information, we knew there was a big gap between the rules that technically everyone has signed off on, and what APPG secretariats understood in practice. This is part of a wider problem where Parliament in principle has rules that in practice are just not strongly enforced. 

For this round we have done the minimal possible request: just asking for membership lists, rather than the wider range of documents we had published previously, and only when both our automated process and volunteers couldn’t find one. Despite this being a relatively clear rule, 94/236 groups didn’t respond to our request for a membership list. 

We also encountered a few groups who did not want to disclose full membership lists for security reasons due to the topic of their group being sensitive, while others were concerned that publishing names could lead to MPs being flooded with unhelpful lobbying. 

We’re sensitive to security concerns and don’t want to strongly argue the point given the small number affected (compared to the much larger number who just didn’t reply), but also there is currently no exemption in the APPG rules for security reasons. If Parliament wants this to be the case, the rules need to be updated to specify the conditions for this exemption from wider transparency.

We will be writing to the Parliamentary Commissioner to report this reasonably high level of non-compliance with transparency requirements of the APPG rules. 

What we discovered

Using the scraper, supported by volunteers’ work, we found memberships for 205 groups online. 

We contacted the remaining groups by email to ask for their membership lists. 140 gave us their membership information, two were in touch but declined to give their lists, and 94 did not respond.

Of the groups we have data for, we found:

• 615 MPs (94%) belong to at least one APPG. Only 35 MPs don’t take part in any at all. This list largely maps onto government ministers, who are not permitted to be a member of an APPG.
• On average, MPs are members of around 10 APPGs.
• Half of MPs are in at least 8 groups, and some are far more active: one MP is listed as belonging to 63 APPGs.

You can view and download the full dataset. 

We also discovered some interesting features about APPGs’ wider memberships  — and that the definition of membership varies between groups. The Guide to Rules states “A member is one who has asked to be on the group’s Membership List” but interpretations of this varied quite extensively. This was especially true about “non-parliamentary membership” (people and organisations affiliated with the APPG, who can be charged for memberships).  Some groups noted that this would include mailing lists with hundreds of individuals so would not share them, while others sent lists of ‘donors’, not all of whom were previously public as they did not meet Parliament’s £1,500 declaration threshold. 

Why it matters

APPG memberships can show what issues MPs care about, and where they might be working across party lines. This matters because of transparency; it’s useful for constituents to know where their MP is spending time and building networks, but also for relationship-building. We think this information can be key to foster common ground both between MPs themselves and between MPs and constituents. 

Explore for yourself

Find your MP’s page on TheyWorkForYou.com or to see their APPG memberships or download the whole dataset. You can also browse this data on the Local Intelligence Hub. Over time, we will make this available on a page per APPG. 

While you’re there, you may spot a few more new features. Join Alex and I on Thursday 23 October for a chatty catch-up on new features, plans for the site, and our vision of a more open Parliament. 

Note: If you are an MP, or on their staff, and our entry is either missing or has incorrect information, you can report issues on this form. 

Photo by Jani Kaasinen on Unsplash

mySociety was founded on one seismic technological change: the arrival of the internet, bringing radical new possibilities to the ways in which we engage with democracy.

Now we’re seeing a second upheaval, just as potentially explosive: the wide adoption of generative AI and machine learning tools — particular kinds of artificial intelligence — not least by the UK government, who have made a commitment to see AI “mainlined into the veins of the nation”. 

From the visible and novel, like ‘AI bot’ MPs; to the hidden and less-interrogated, like the algorithms that drive decision-making around benefits; to the new capabilities around working with large text datasets that we ourselves are experimenting with at mySociety: artificial intelligence is changing the way democracy works. 

We’ve been thinking about AI for some time, as have our colleagues around the world — TICTeC 2025 had a strong strand of pro-democracy organisations showcasing how they are using new technologies to hold authorities to account and support public engagement; alongside developers showing the tools that aim to make the government more responsive. 

AI is coming to democracy, whether we like it or not. In many places, it’s already here.

But there are implementations in which it can be highly beneficial to us all; and ways in which it can present a clear and present danger to democracy. 

It benefits everyone if there is a high level of understanding of both the challenges and the opportunities of AI in government. Democratic decision makers need to understand digital tech in order to legislate effectively around it, to develop and procure it effectively.

This is not just so that they can deliver services more efficiently, but also to ensure that they retain the legitimacy of democratic government by using tech and AI in a way that ensures transparency and accountability, preserves public trust and allows the public to understand and participate in the decisions that affect their lives. 

Reflections for our time

Over the next few months, we’ll be sharing our own thoughts and experience — alongside invited guest writers who are thinking about how AI interacts with democratic processes and institutions, and how to make that better — in a series of short pieces.

These will examine the different ways that AI is affecting the things we care about here at mySociety: 

• Transparent, informed, responsive democratic institutions 
• Politicians and public servants who work for the public interest 
• Democratic equality for citizens: equal access to information, representation and voice
• A flourishing civil society ecosystem
• The effective and principled use of digital technologies
• Action from politicians to match the evidence of the climate crisis and the level of public concern 
• Better communication between politicians and the public, creating space for climate action. 

Stay informed

If you’d like to get updates in your inbox, make sure you’ve checked ‘artificial intelligence’ as an interest on our newsletter sign-up form (if you already receive our newsletters, don’t worry – so long as you use the same email address, this will just update your preferences. Just make sure you’ve ticked everything you’re interested in).

By also completing the ‘how do you identify yourself’ section, you’ll help us send you the most relevant material: that means guidance if you work in government or build tech; data and our analysis if you’re a researcher; tools for holding authorities to account if you are an individual or work in civil society, and so on.

—

Image: Adi Goldstein

Once again, the TICTeC Communities of Practice have given us all the opportunity to learn from those at the frontline of civic tech: this week’s session, Beyond websites: how pro-democracy projects reach their audiences saw practitioners from Georgia, Nigeria and Uganda explaining the ways in which you can engage with audiences beyond a website.

You can watch the session here.

Ana Arevadze from ForSet explained the care and attention that the organisation put into making sure that an election education campaign, delivered by influencers, was a learning experience for all involved. This was a presentation that a small group of people had been fortunate to experience at the ATI Day in Mechelen, but is now available for all to watch. 

Ufuoma Oghuwu from Enough is Enough Nigeria outlined how the Shine Your Eye website provides citizens with information about their elected officials — something that’s often missing after the canvassing and electoral cycle has passed — and then described how that information has a life beyond the website, thanks to chatbots, WhatsApp and social media.

Last but not least, Joseph Tahinduka of ParliamentWatch Uganda shared the fantastic efforts they go to to make parliamentary activity accessible to the social media generation, who so greatly prefer short videos to trawling through lengthy reports. Is it time for all of us to start getting onto TikTok? You’ll have to watch to find out!

Sit back and enjoy the video: there was so much to learn from our speakers, and we’re glad to be able to share their insights with our networks. 

Our WriteToThem website makes it easy for anyone to send an email to their elected representatives. That’s the core concept, and it works brilliantly for millions of users every year — but that said, we’re aware that even when a website is simple and built with usability at its core, not everyone has an equal ability to access it.

As part of our warm up for a new programme of work on WriteToThem, we’ve published some of our internal research from a few years ago on messages written on behalf of someone else — what we’re calling ‘proxy use’. 

The reasons for this are easy to understand: the primary subject may not be confident at writing in English; may be elderly or have a condition that makes it easier to delegate the task of writing; or may generally use internet services through intermediaries.

The key findings are:

• A small group of users (5%) were writing on behalf of someone else.
• Proxy messages made up 6.8% of messages to local councillors, and 4.5% of messages to MPs. This would account for an estimated 55,000 messages to MPs through the service in 2019.
• The largest group was people who were writing on behalf of family (40%), but there were also people writing on behalf of local groups (40%), friends or people they knew (12%), and service providers writing to representatives on behalf of clients (8%). Messages on behalf of clients from carers would have accounted for an estimated 7,500 messages in 2019.

We’re about to embark on research and development work around WriteToThem, and these findings will contribute to our understanding around making it easier to get the right type of message to the right place.

If you are interested to dig deeper into everything we discovered around proxy usage, take a look at the full piece of research here.

Thanks to new funding from the Welsh Government’s Democratic Engagement Grant, we’re going to be doing some really exciting work around WriteToThem over the next couple of years, specifically focused on helping people in Wales. This grant is both an opportunity for us to improve our approach, and to help get our tools into the hands of people who can take it further. 

WriteToThem’s core mission is to make it very easy to send an email to your politicians — and to help people send the right message to the right representative. This is especially relevant in Wales, where devolution brings decision-making closer to people, but can also mean people have to discover who is responsible for different public services. 

And so, what does this new bundle of work look like?

• First off we’ll be doing some concentrated research with representatives and community groups to understand barriers to constructive communication, which we’ll use to inform new development on the site.
• We’ll also be doing work to ensure that more people are supported to write to representatives for the first time, particularly in the most deprived areas of Wales, where typically there is less engagement with democracy — and all the more need for it!
• Just as important will be the translation of every part of the WriteToThem user flow into Welsh — that’s webpages, buttons, confirmation emails, error messages, the lot. 
• And finally, there’ll be improved guidance about where to send messages — people already appreciate WriteToThem for its succinct descriptions here, but we know there are improvements that can be made, especially in the devolved regions.

We’re excited to get going on this, and to work with other grantees on how we can help each other go further. We’re starting to plan our research phase, and will have more to say about our plans soon, so watch this space or sign up to our newsletter to be the first to know.

 —

Image: Catrin Ellis

I’ve written before about how we’re thinking about “low resource” use of Large Language Models (LLMs) — and where some of the benefits of LLMs can be captured without entering the “dependent on external API” vs “need new infrastructure to run internally” trade-offs.

One of the use cases we have for LLMs is categorisation: across parliamentary data in TheyWorkForYou, and FOI data in WhatDoTheyKnow we have a lot of unstructured text that it would be useful to assign structured labels to, for either public facing or internal processes.

This blog post is a write up of an experiment (working title: RuleBox) that uses LLMs to create classification rules, which can then be run on traditional computing infrastructure without dependence on external APIs. This allows large-scale text categorisation to run quickly and cheaply on traditional hardware without ongoing API dependencies.

Categorising Early Day Motions

We have a big dataset of parliamentary Early Day Motions (EDMs), which are formally ‘draft motions’ for parliamentary discussion but effectively work as an internal petition tool where MPs can signal their interest or support in different areas.

For our tools like the Local Intelligence Hub (LIH) we highlight a few EDMs as relevant to indicating if an MP has a special interest in an area of climate/environmental work. We want to keep these up to date better, and to have a pipeline that’s flexible for future versions of the LIH that might focus on different sectors. We want to be able to tag existing and new EDMs depending if they relate to climate/environmental matters, or other domains of interest.

A very simple approach would just be to plug into the OpenAI API and store some categories each day, but this is giving us a dependency and ongoing cost. What we’ve experimented with instead is an approach where we use the OpenAI API to bootstrap a process. We’ve used the commercial LLM to add categories to a limited set of data, and then seen how we can use that to create rules to categorise the rest.

Machine learning and text classification

Regular expressions and text processing rules

The “traditional” way of classifying lots of text automatically is to use text matching or regular expressions.

Regular expressions are a special format for defining when a set of text matches a pattern (which might be “contains one of these words” or “find the thing that is structured like an email address”).

The advantage of this approach is that you can see the rules you’ve added and at this point the underlying technical implementations are really fast. The disadvantage is that you might need to add a lot of edge cases manually, and regular expression syntax is not always clear to understand.

Machine learning

The use of “normal” machine learning provides a new tool. Here, models that have already been trained on a big dataset of the language are then fine-tuned to map input texts to provided categories.

The theory of what is happening here is that in order to accurately “predict the next word”, language models need to have developed internal structures that map to different flows and structures in the text. As such, if you cut off the final “predicting the next word” step, and replace it with a “what category” step, those internal structures can be usefully repurposed to this task.

As such, machine learning based text classifiers can be more flexible. They are picking up patterns like “this flavour of word is in proximity to this flavour of word” that would be difficult to manually code for. The downside is that they are a black box, and it is hard to understand what it has done to make a classification decision. They are also more resource intensive and slower to categorise large datasets — but still fundamentally possible to run on traditional hardware.

LLMs

The next wave is LLMs, which take the same basic concept and massively increase the data and the size of the model. Here, rather than replacing the “next word” step, the LLM is trained on a datasets that contain both instructions and the results of following those instructions. This makes zero-shot classification possible. Without retraining, a model can be given a text and a list of labels and it can assign the label.

This remains a (now massive) black box, but errors in category assignment can be improved by adjusting the instructions. The new downsides over smaller machine learning models is the much larger size of the model hugely increases the cost of self-hosting and creates dependencies on external companies providing models. If you use proprietary models (that are regularly updated and deprecated) this creates problems for reproducible processes.

Rulebox approach

The Rulebox approach combines aspects of both approaches. One of the things that LLMs are quite good at is writing code to solve stated problems. Here we’re doing a version of that: providing text and a category, and asking it to produce a set of regular expressions that should assign this category.

This has its unique set of pros and cons: you are still bound by the underlying problem of regular expressions that they are matching on text rather than the vibes of the text (which language models are better at). But you have massively reduced the labour time needed to create the huge set of rules, and once you have these they can be applied at speed on traditional hardware.

This is part of a focus on “low resource” use of LLMs – where we want to think about where we can get the most value out of new technology, in a way that avoids dependence or hugely increased capacity.

The process

We used an OpenAI-based process to assign labels to a set of 2,000 EDMs (1000 each for a training and validation dataset).

We then created a basic structure for holding regular expression rules using Pydantic for the underlying data structure of the collection of rules. For each rule, this either allows a list of regex expressions that are AND (all must match) or OR (one must match) — with the option of NOT rules that will negate a positive match.

Once we have the holder for a set of rules, and a dataset with a set of labels, we can start to calculate mismatches between what the rules say, and the result. Running this in a loop with steps that query an LLM helps refine the result.

The steps are:

• Calculate mismatches between ground truth labels, and assigned labels: finding both missing labels and incorrect labels.
  • AI: For each missing label, create a new regex rule that would assign the correct label.
  • AI: For each incorrect label, adjust and replace the regex rules that triggered this label.
• Repeat until no missing or incorrect labels.

PydanticAI is used to interface with the OpenAI API. This includes not just using pydantic to validate the returning data structure, but some extra validation rules that the resulting rules match the text that was being input. So for instance, if a rule is being generated to assign a label to a piece of text, if the generated rule fails to match the input text, this failure is passed back to trigger a retry.

The initial attempt at this got stuck in a loop creating rules that were too general, and trying to narrow them down. At this point, we cut the categories down to just a few we were really interested in, and after that performed better, expanded out to eight where it felt like keyword categories should perform reasonably well (or at least successfully generate rules). This ends up with 1,500 regular expressions to assign eight categories.

Applying the rules

Once we have the rules, we know they work for the training dataset, but how useful are they in general?

Using the validation dataset, we can see the following differences:

• Correct labels: 230
• Missing labels: 73
• Incorrect labels: 41
• Items where no labels were assigned: 808 / 1000 total items

Reviewing these, incorrect labels generally felt fair enough – these tended to be examples that contained obvious keywords related to the environment, but were part of longer lists where the labelling process did not judge it as one of the focuses of the text. The missing labels were more of a problem, where 33 of the missing labels were environmental ones. Expanding the training data should improve this, but there is always just going to be a long tail that’s missed.

Something else we experimented with at this stage was moving the process that applied the rules from Python to Rust (using an LLM to translate a basic version of the Python mechanics). This cut the time taken to categorising 13,000 EDMs from 2 minutes to 4 seconds. The benefit of this isn’t just being fast on this dataset, but that much more complicated rulesets would not be a big slowdown.

What have we learned

In general this is an approach worth investigating further as a bridge between several useful features: with it, we are able to translate an initial high intensity of LLM into a process that can be run fast on traditional hardware, and importantly is not a black box in terms of how it assigns labels.

It doesn’t completely carry over the benefits of LLMs:it is better for smaller, more precise categories. It really needs a good theory on why a keyword approach would be a good way of categorising something. It might be a good transitional approach for a few years while options stabilise around more open models with lower resource requirements.

Next steps

The next steps on this are to expand the training data a bit and start seeing if we can practically make use of the categories assigned, or if the accuracy causes problems.

Depending how this goes, we can revisit the initial experiment code and tidy it up into a more general classifying tool. This could tackle other classification problems we have that might be suitable, and we could make the tool more widely available. An advantage of this kind of approach (as our previous work around vector search) is it is the kind of project where “a technically-minded volunteer helped us to create a tool” might help organisations without creating significant new dependencies or new infrastructure requirements.

We also want to think about where hybrid approaches might be useful. For instance, in these datasets, most items are not labelled at all. A fast first pass that identifies potential items could then switch to an LLM approach to knock out false positives from the data. Similarly, once we have a smaller pool of environmentally-linked items, further subclassification using LLMs is much more viable.

Our general approach is to try and identify the things that LLMs can do uniquely well, and build them into overall processes that tame some of the things that worry us about AI in general. Here we are exploring how we have focused on the use of LLMs, resulting in new processes that are both fast and efficient. For more about our approach, read our AI framework.

Photo by Marc Sendra Martorell on Unsplash

Transforming and publishing official Parliamentary transcripts is one of the key activities of parliamentary monitoring organisations (PMOs)  — in our case, that means running our website TheyWorkForYou, but there are many organisations around the world doing the same for their own parliaments. 

Building on top of transcripts means that PMOs can focus their time on where they can add value to those transcripts: either applying their digital skills to make them more accessible, or merging them with other datasets and sources in ways the official Parliament sites cannot. 

One of the interesting things about Parliamentary transcripts is that they don’t exactly reflect what happened. They can be an official record of what was supposed to have been said rather than what was said: a constructed version of Parliament that is close to, but not exactly, reality. 

This can be very important when democratic needs are not for verbatim transcripts. MPs can ask to make corrections on factual content if they misspoke. The Record can add useful shorthand, referring to standing orders, or additional information that was not said orally. For parliaments with multiple official languages (in the UK’s case the Senedd/Welsh Parliament), transcripts make parliamentary activity accessible in all official languages. 

But the difference can also be political in the “first draft of history” sense. The transcript can retrospectively apply rule-following in a way that can remove political speech on the edge of those rules. We’ve noted before there are times where the transcript does not reflect reality in procedurally significant ways. Historically famous events can be at the edge of the transcript, rather than visible in it. 

So there are pros and cons to our understanding of parliamentary activity being dependent on transcripts alone. But the existence of accessible video recordings of Parliament makes it easier than ever to see the difference between what officially happened and what actually happened.

Combining transcript and video

In 2008 we did some work with the BBC to explore how to make parliamentary video searchable. At the time the best approach we had was a crowdsourcing approach to reconcile timestamps and speeches  — around 400 volunteers aligned 160k speeches. Ultimately this process had big technical overhead in video storage, and a lot of manual work was required to keep the two feeds together, and we stopped using video in this way. 

New approaches make this possible at scale with much less manual effort. As part of our TICTeC Community of Practice around Parliamentary Monitoring we ran a session on video and transcripts, hearing from OpenParliament.tv about their approach (video of the presentations). 

OpenParliament.tv currently covers the German Bundestag, but is interested in expanding the approach to more countries. The platform is a combined video/transcript search platform, where individual speeches can be searched for, and jump the video to that point in the record. 

There’s a lot of moving parts that make this work, but the core work is in how the video and the transcript are aligned. The transcript is converted to computer-generated audio, and then the generated audio is matched against the real audio. The matching uses an adapted version of the aeneas framework for ‘forced alignment’ of text and speech. 

When you think about it, this approach makes a lot of sense: speech to text is often specifically bad at generating the punctuation of written language, while one of the key things in syncing transcripts to video is finding the start and end of blocks. This can still run into difficulties when what is being said is just not present in the transcript, but generally it can flow around problems and match items on either side. From an international perspective, this is also interesting in that it’s an approach that works better across different languages than speech-to-text approaches. 

In other technical details, Open Parliament TV does not host the video themselves. Offline they need to process it to extract and match audio, but online their player links to the videos as hosted by the Parliament. In the long term, they want a workflow that sends videos to an internet archive as a backup. This is another useful purpose of democratic transparency project: – to act as civic redundancy against backsliding access to democratic materials. 

Switching between modes

We had a bit of a discussion in the group about transcripts, and about where AI approaches might make it possible to generate them. This might be useful in settings where there is only video output, to make it easier to search and parse. This could also be a useful bridge in cases where there is a significant delay before transcripts are released. 

But official transcripts are an art beyond just writing down what’s happened,  imposing a consistency on the parliamentary record that is very useful as a building block connecting it to other data. For instance, Open Parliament TV does additional detection of named entities based on the transcript, meaning that specific mentions can then be seen in the video. 

The future of parliamentary monitoring might be switching between these modes: making the ground truth of what happened visible through video, augmenting this with the formal transcript, and bridging from that to other sources of information. In short, pulling on what different mediums do best to make democratic processes stronger and more transparent.

Header image: Photo by Diego González on Unsplash

We don’t talk a lot about elections at mySociety: we see our unique contribution as being about the democracy between elections, with a focus on how democratic institutions work (sometimes how they can work better), and the connections between the public and those institutions.

But whatever part of the democratic system you care about, elections shape everything. Rules about elections not only decide the winners, but the incentives that all players operate by. In our work looking at money in politics, we started looking at broken forms and worked our way back to problems in how elections are financed. There’s a lot of good work to be done on small problems, but we need to keep our eye on the big picture too.

This blog post looks at the new government’s new election strategy (which will be the basis of the forthcoming Elections Bill) 

This strategy provides some certainty around large-scale changes in how UK elections will work in this Parliament, which is good as we’re at the point in the election cycle when the practical work to make these changes happen needs to begin.

The strategy is a solid foundation to build on, that needs to be followed up with complementary support and legislative action through the rest of the Parliament. There are also areas where the plan doesn’t go far enough – and from the outside we need to be building better evidence and campaigns for change. 

Making it easier to use your vote

The big headline changes are:

• automatic voter registration
• expansion of valid Voter ID
• votes at 16.

These are all changes that make it easier for people to vote if they want to. 

It is important that these changes are enabled now to facilitate the substantial work that will take years — both within the electoral administration system, and for civil society. 

One of the significant hopes for votes at 16 is that it will give more people a first election while they’re in an educational setting that helps them understand and use their new right.  But this doesn’t happen on its own: the Democracy Classroom have outlined the work that is required over the next few years to create the environment where young people not only have the right to vote, but the knowledge and willingness to use it. 

Election finance and donor caps

There are good foundations on improved electoral finance laws in the strategy.  A key improvement is much stronger regulation of unincorporated associations: creating Know Your Donor requirements, and bringing the transparency thresholds into line with donations to parties and candidates. These transparency thresholds are too high (and have gotten higher), but this is an area that could (and should) be redressed through a statutory instrument at a later point.

What this strategy does not go near is the idea of donation caps to limit the power of large donors. As we explored on our Beyond Transparency report, a key blocker in this area is both the comparative lack of public funding of elections in the UK compared to European and Anglophone countries, and a belief that public opinion is an obstacle to a move in that direction (public funding is not popular on its own, but the status quo is also deeply unpopular). 

In the spirit of “what is the work needed now to enable change by the end of the Parliament”, our key recommendation is that either government or civil society actors convene a citizens’ forum/assembly to further unblock this line of argument by creating a better understanding of how the public approach trade-offs in this area. There is good reason to believe that, while exactly what it looks like might be open to debate, increased levels of public funding are possible, enabling donor caps and restricting the uneven influence of wealth on elections. 

In the absence of this, we are unlikely to make substantial progress within this parliament. We need well-developed answers to the “how are we going to pay for elections then?” question  — and to be building those answers now. 

The Electoral Commission needs better data infrastructure

A key part of the new strategy is creating much greater enforcement powers for the Electoral Commission, increasing maximum fines and expanding to cover financial offences by candidates and local third-party campaigners. 

This implies more resourcing for the Electoral Commission, but also improved data infrastructure.

For instance, basic infrastructure (such as the Electoral Commission having a list of all candidates covered) does not exist in a formal centralised way. This is currently provided indirectly through Democracy Club’s work using volunteers to source hundreds of different statements. 

The strategy provides both a need for a better solution, and the framework of how it can happen – through new requirements for Electoral Register officers to provide election information to the government and Electoral Commission. Building on this legal foundation will require important work over the next few years to create reliable flows of data to enable both better public participation in elections, and an effective regulator of candidates and parties. 

For investigation of campaign finance offences,  the flow of financial information currently has significant problems, not only in terms of external accessibility but in terms of usable data for a regulator to enforce the rules. Our forthcoming report, Leaky Pipes, will explore this problem in more depth, and propose solutions that help support both public transparency and effective regulation. 

Abuse and intimidation

The strategy includes a number of elements aimed at reducing the incidence and impact of abuse and intimidation of candidates and election staff.  These range from privacy measures such as removing the need for candidates’ home addresses to be published, through making intimidation related to an election an aggregating factor in other offences, and requiring candidate ID checks to avoid sham candidates. 

A key aspect of this in the strategy is better guidance on the different parts of the system and creating clearer understanding on roles, available protection, thresholds for action and ways of accessing police support. 

A throughline of the Jo Cox Civility Commission report is the potential high rewards of better joining up and signposting existing systems of support.  A consequence of creating the guidance may be discovering further opportunities for better joined-up procedures. 

This is another area where improved data infrastructure would enable a range of goals, for instance making it possible for consistent updates to guidance to be sent centrally from the Electoral Commission.

Managing scope

A running theme through the strategy is trying to close loopholes that can be abused. However, some loopholes reflect genuine ambiguity that can be hard to address with either creating a chilling effect or problems with enforcement through covering a large number of organisations.  One area of concern is the expansion of the need for a digital imprint to cover unregistered campaigners (the name and address of the organisation creating/promoting a viral post), which might bring a huge number of organisations into scope. We would want to see more examination of this as the bill progresses through Parliament. 

Character and truth

One interesting element of the strategy is support for the recommendation that the current Speaker’s Conference on the security of MPs, candidates and elections establish a code of conduct for campaigning and coordinating cross-party discussions. 

Given the Speaker’s Conference‘s specific interest in s106 (the rule where defaming a candidate’s character can be an election-invalidating offence), discussion of conduct inevitably enters grounds around truthfulness. The problems in agreeing a code of conduct was one of the reasons that the Advertising Standards Agency stopped policing election adverts in 1999 (the lack of regulation here is generally not known by the public). 

Progress here could work as an enabling measure for the ASA to adopt New Zealand style rules on policing political ads. But it could also run into problems finding consensus. It is worth campaigners in this area paying attention to where practical progress is possible, and could be enabled by outside research and campaigning. 

Electoral Commission independence remains at risk

This strategy intends to use the government’s power to set the Electoral Commission’s strategy and policy rather than abolish that power. Labour was opposed to this in opposition. It undermines all electoral offences if parties have a viable “win at any cost” approach to electoral compliance and then can deprioritise enforcement of rules against them once in government. 

As the Chair of the Electoral Commission said, they “remain opposed to the principle of a strategy and policy statement, by which a government can guide our work. The independence and impartiality of an electoral commission must be clear for voters and campaigners to see, and this form of influence from a government is inconsistent with that role.”

This is a change that needs to be made through primary legislation. Failing to do this within this parliament will set a norm where two major parties have agreed that electoral enforcement priorities are set by the winner, rather than reflecting an agreed democratic understanding of the rules. 

Improving candidate/voter privacy

In the 19th century, when the franchise was heavily restricted, the list of electors was public so that people’s eligibility could be challenged. This has continued into a big dataset of all registered electors being commercially available to buy from local authorities. 

This is out of sync with modern ideas of privacy, and especially the risks of these kinds of massive datasets. The opt-out to the open register was added in 2002, and by December 2018 56% of the register had opted out. The usefulness as a universal dataset has now been broken, and it serves little purpose. 

The election strategy says that for automatic voter registration, opt-out will become the default. We would recommend going further to fully remove the register.

This is just one way that electoral transparency is out of step with our current democratic needs. In our forthcoming Leaky Pipes report, we explore how GDPR is both used as an obstacle to useful donor transparency, while at the same time a surprising amount of information on small donors is available for public access.

—

Image: Ben Allen

We have completed our quarterly update to the TheyWorkForYou voting summaries and they’re now complete as of the end of March 2025.

We’ve added 20 votes to TheyWorkForYou’s voting summaries, covering the first three months of 2025. We’ve also added several votes from the 2019-2024 parliament retrospectively when creating a voting policy in a new area. 

To learn more about our process, please read our previous blog post.  We have also recently released TheyWorkForYou Votes which, as well as providing open data for anyone to use in their own online parliamentary projects,  is  powering TheyWorkForYou’s voting summaries. 

This update has added new votes to existing policies:

• Climate change
• Low carbon electricity generation
• Smoking bans
• Increased capital gains tax
• Windfall oil and gas tax
• Employment rights
• Assisted dying (see note below)

We have also added four new policy lines:

• Renters’ rights
• Charge VAT on private school fees
• More powers to investigate welfare fraud, including requiring banks to monitor accounts of welfare recipients
• Nationalising teacher pay and the curriculum for academies, tightening child protection duties, free breakfast clubs (Children’s Wellbeing and Schools Bill). More about this below.

And behind the scenes, we’ve created four policy lines aren’t currently live, but could be used if and when we add historic or future votes on these topics:

• Greater alignment of UK product standards/measures with EU standards 
• Speeding up nationwide infrastructure consents and opposing local veto powers over large energy schemes 
• Draft border security bill policy
• Measures to encourage purchase and use of electric vehicles

Notes

Assisted dying

In our previous blog post we flagged that we might treat the third reading of the Terminally Ill Adults (End of Life) Bill differently, outside our normal process. This policy has been updated ahead of when it would appear in our normal review. 

We are now scoring based on the third reading rather than the second reading. Normally, there is almost no difference between these so we include both to cover people who are absent from one. 

But when people change their minds (which is more likely on a free vote), we should prefer the later vote (otherwise we create ‘voted for and against’ lines when the final one is clearer and more meaningful). 

As 58% rather than 62% are now in favour of the motion within the Labour Party, it no longer counts as a significant break from the party for those who voted in the minority. 

We also as a result of this have:

• We have amended some of our copy on significant differences and the whip to provide more recognition that significant differences between an MP and a party can emerge in a free vote. There was a fair comment from an MP that previously it could be inferred that this vote was whipped from the two bits of text close together (which is the opposite of the complaint we were trying to avoid here, that the summary implies there is no whip).
• added a new ‘tended to vote for/against’ label to better describe situations where a party is very split (these are rare, but worth having the language for). 
• added the full percentage of the alignment score for the (few) policies that fall between 35-65% policy alignment – to give a bit more information on the direction of the lean. 

Votes covering multiple topics

The TheyWorkForYou voting summary model works best when over time there are a range of votes on essentially the same principle. Assisted dying is a good example of this. 

Where the model struggles more is when a vote covers multiple principles. Sometimes you get a big vote that is clearly about one thing,  but often a bill does a range of stuff (A, B, C). 

You can do a few things here. You can say it mostly does one thing (other items less notable). You can also create a range of policies (voted for A, voted for B, voted for C) to reflect what happened. But if you then take this in isolation, it loses that context of “this was a vote against A as well as against B”.

Ideally, if there’s a simple thing, we should use that — but when it can’t be squeezed down to something simple (and especially when the line of opposition critique is not necessarily the main thing the bill does) we should try and reflect that in the simplest way possible. 

We’ve recently added the ability to view the voting records only for the tenure of specific governments. This gives us a bit more flexibility to add policies without being crowded by old votes,and lets us include policies that are very time specific and unlikely to be updated.

So for some votes, the policy description will effectively cover only that bill. For example:

[x] voted for nationalising teacher pay and the curriculum for academies, tightening child-protection duties, free breakfast clubs (Children’s Wellbeing and Schools Bill).

Ideally these would be the exception rather than the rule, but we have to reflect what is actually happening in Parliament, while trying to be both clear and accurate in how we present it. 

Electric vehicles

We created this policy because The Motor Vehicles (Driving Licences) (Amendment) (No. 2) Regulations 2025 flagged that there had been a run of relevant legislation on this issue. 

The Draft Vehicle Emissions Trading Schemes Order 2023 is included retrospectively.  We have also included an agreement for the approval of the ‘Public Charge Point Regulations 2023’. 

Finance bills over the last five years have been relevant to this,creating incentives for building  charging infrastructure through an 100% first year allowance, and renewing this each year. But then you also run into the 2024 Finance Act, which as well as extending this, started charging vehicle excise duty (VED) for zero emission vehicles (so could logically be counted as both encouraging and discouraging purchase and use of electric vehicles ). 

By the ‘cohesion’ element of our criteria we want to avoid too much use of votes on many items (like finance bills) in policies on a single item. In this case, we do not want the policy to be mostly about votes on finance bills given it is the same measure being extended multiple times. As such, we are including the initial 2021 introduction of the policy only. Other finance bills are included as informative votes that do not contribute to scoring. 

We have not included an agreement for ‘Electric Vehicles (Smart Charge Points) Regulations 2021’ for technical reasons (it was approved in a run of Statutory Instruments that have not been picked up well by our agreement detector). This is a candidate for inclusion when adding the policy in future. 

Gambling and agreements

A note on something we’re choosing not to include at this stage — but we’re open to feedback. 

The Gambling Levy Regulations 2025 was a statutory agreement adopted by agreement with no opposition (but when it was discussed in committee there was opposition). 

There are few options on how we could treat this. If it was a division, we would likely create a new policy on a gambling levy and assign it as a scoring vote (we could also extend our existing gambling regulation policy). 

However, as it’s taken by agreement (with no individual votes), there are a few possible approaches. There’s an “it’s not a vote” view, which means the summaries should stay focused on votes which have a common interpretation (if not reflecting the MP’s personal views, at least demonstrating that they did show up and vote for/against something). 

But then there’s the “impact is what matters” view, which is that it’s perfectly correct to include decisions like this on all MPs’ voting summaries as a record of impact (and to exclude it is to systematically miss points of consensus). That people might or might not support it in their hearts is interesting to understand, but ultimately irrelevant. 

Here is what we said in our 2024 review about the trade-off:

In general, our starting point should be accurately describing what is currently happening. “Agreements” are a large part of the current picture of how Parliamentary power works. We want to include decisions without a vote as a way power is exercised, while being accurate in what that reflects about individual MPs, and with an eye on making a point about the scale and lack of scrutiny of secondary legislation.

As such, we are starting cautiously. We have built technical approaches that let us include references to agreements in scoring and informative roles in a policy. We are applying this in a limited sense retrospectively, and will apply the same criteria used for vote inclusion to agreements moving forward – but may for the moment prefer not to include for borderline cases.

We have previously included agreements as components in policies that mostly contain votes (although depending on an MP’s tenure, agreements might make up more of their score).  The caution here is that this would be a one item policy, and so a step beyond how we’ve used it before (where it supplemented other votes).  As such, we are leaving this one in a draft policy for the moment, and might revisit on either future votes, or in an end of parliament retrospective. 

Greater UK/EU alignment

This is a draft policy to capture the Product Regulation and Metrology Bill. We have not made it live because logically there are some votes in the last five years that should also be included in this, and we have not had time for a comprehensive review. 

Anything we’ve missed?

We have a reporting form to highlight votes that should be added/are incorrectly in a policy, or a substantial policy line we are missing. We review responses for urgent problems, and otherwise these comments feed into the periodic updates.

What else we’ve been working on

We recently released TheyWorkForYou Votes, the backend we use to power the voting summaries for public use. You can watch our launch event for more about the wider context of the work. 

We’re continuing to do work around MPs’ financial interests: you can learn more on our WhoFundsThem page. 

Support our work

We create the voting summaries because we think it’s important to keep track of (and make more visible) decisions made by our elected representatives. We take doing getting this right seriously, and want to give creating the summaries the time they deserve.

If you value the work we do around voting records and would like to support our work, please consider donating. 

—

Image: Engin Akyurt