Clara Yeo

Given the plethora of highly interesting ideas, projects, domains, and disciplines that our rich world is built of, how can one choose what to work on?

There is an argument in favour of long-term projects (missions): there are some affairs in the world today that seem to be solvable only with substantial persistence and patience. These issues, which are both limited to a single discipline but more frequently across multiple, seem to require considerable multi-year commitment before meaningful progress can be achieved.

Therefore: the question of what to work on flips: it's not only what one wants to work on but also what needs to be worked. Yet, one is unlikely to devote one's life to something merely because it is needed, without also finding it compelling. We are interested in the intersection.

How does one choose a mission if they decide they want to work on one? The answer is nontrivial because of the requirements:

(a) it has to be something that is needed in the world -- (usefulness dimension)
(b) it has to be something in which they have above-average competence (only insofar as the good outcome is realised) -- (aptitude dimension)
(c) it has to be something they want to work on -- (preference dimension)
(d) it has to be something they want to work on long-term — a successful mission becomes one’s life’s work -- (commitment dimension)

The difficulty of the answer lies in the inexact comparison of the degree of each of these dimensions per issue. For example, here's a list of issues:

1. Alzheimer’s
2. Economic stability of Western countries
3. Mental health of the young population

Let's reflect on each dimension.

(a) Usefulness: they are all extremely important; maybe one could claim economic stability is more important due to its impact to much more of the population of a country but that wouldn't be a convincing claim
(b) Competency: judging one's competency for each of these can be quite hard insofar as what skills are required for each problem/solution is genuinely unclear (even if a problem seems medical or economic, the practical path to the solution can require enterpreneurship or campaigning skills)
(c) Preference: people are interested in more than one disciplines. Helping people with medical problems can be fascinating and rewarding but so can scheming plans for eradicating homelessness.
(d) Commitment: all of these issues seem so distant and commitment a fairy tale unless one has already spent some time on them

I think this demonstrates the heart of the problem of choosing a mission. There are convincing arguments that each of the dimensions can be both very low and very high for someone. It's just very hard to say! So what do we do? I think our past experience is the best place to start.

1. Which country-level or global scale problem keeps coming back to your mind?

2. Is there a discipline or skill you feel you have a natural talent for?

3. What is more likely to work out depending on who you know and where you are? Try not to be biased when answering this. The point is the good outcome for the world! (For example, don't feel guilty if you are the child of the prime minister or something; whatever power you have, use it for the benefit of all)

4. What issue can really grab you? What do you see could be your life's work?

I feel these are good thinking paths to reflect and understand such a mission project. I've already been verbose but here are some more aspects:

• If one feels their reflections yield little insight maybe there is a lack of experience. So try new things! Experiment, do things you wouldn't normally do!
• It's worth thinking about one's mission intentionally. Find a weekend, clear your schedule, go on a walk, and just think about this — no other expectations other than spending time on thinking.
• Choosing is better than not choosing, even if the choice ends up being suboptimal. That's not to say one should rush a choice. Take a few years to think about this, experiment, try things out. But I would argue in favour of choosing a mission, eventually.
• I am not saying everyone should have a mission. Such an approach to a career/life is not for everyone. But I think it is for the people that are here at Newspeak House.

Last year I teamed up with Georgia Bullen, Executive Director of Superbloom Design, to offer the cohort a session on Funding 101 for Civic & Political Tech. We covered a lot of ground in just an hour, including candid takes on the funding landscape, entity and legal eligibility types, and winning strategies. Now I work for a foundation, so hopefully it'll be even better!

A casual, ask-the-expert session on how projects actually get funded in the real world with Georgia Bullen and Matt Stempeck.

​If you’ve ever thought “I have an idea… now what?”, this is for you.

​We’ll discuss the practical side of getting a project off the ground: where funding comes from, how to find opportunities, what early-stage teams should prioritize, and what makes an application stand out when you don’t yet have a track record.

​This session is especially relevant for fellow candidates thinking about what happens after a prototype: turning something scrappy into something sustainable.

​Topics we’ll likely touch on:

• ​Where to find funders and active opportunities

• ​What first-time projects should look for (and avoid)

• ​Legal structures and setup basics

• ​Signals of credibility when you’re unknown

• ​Partner development and early traction

• ​What strong grant applications actually look like

​Georgia is the Executive Director of Superbloom, where she works at the intersection of technology, movements, and public interest infrastructure, supporting teams building tools for civic participation, accountability, and collective action. Her work spans fundraising strategy, ecosystem building, and helping early-stage projects navigate the path from idea to impact.

​Matt Stempeck is a civic technologist and researcher focused on how technology can strengthen democracy and support real-world civic action. He is the founder of the Civic Tech Field Guide, a global collection of tools and organizations working in the public interest. He is faculty at Newspeak House (the London College of Political Technology) and has held roles across government, academia, and industry, including at Microsoft and Cornell Tech. His work focuses on helping projects move from ideas to impact, with an emphasis on practical strategies for participation, funding, and movement-building.

Collective intelligence as governance of AI, and AI for collective intelligence.

This topic covers the distinction between democracy and collective intelligence, why you might choose one over the other, and where they trade off. How does human CI and artificial CI intersect? What can societies of humans and agents can produce? What are the open questions?

Some examples

• Democratic Inputs to AI - CIP & Open AI
• Collective Constitutional AI: Aligning a language model with public input - CIP & Anthropic
• Community Models
• weval.org & participatory evaluations

On human & agent societies

• The Habermas Machine & habermolt - where next?
• Harnessing the Collective Intelligence of AI Agents in the Wild for New Discoveries

Kyoto is a play about the negotiations leading up to the Kyoto Protocol, including the UNFCCC which was the world's first international accord on climate change. It can be purchased and read in book form.

I'm about halfway through as of posting this, and it's a really interesting demonstration of negotiation. I'm reading it through the lens of standard MBA negotiation theory, and would love to discuss with others in that context. Like a one-off book club.

So it turns out, in government, civic society, and life more generally, finding out of if something actually works (eg, achieves the change you wanted it to) is actually rather hard. People have done excellent talks about this. The UK Government has got a book about this. And yet it's still all pretty hard, and people tend to be very confidently wrong.

I've spent a lot of my time trying to run evaluations in government, but it's not exactly a fun thing to teach... so instead, I'm proposing a bit of a lab experiment: let's run an experiment on ration club (or something else lab related you control!)

If people are interested, we'll spend a few hours talking through what sort of changes we might want to make to RC, then hopefully actually test them, robustly, record what changes, and run a full experiment from cradle to grave. We'll start from first principles, so a vague understanding of causal inference and statistics is very welcome, but absolutely not necessary.

PS: it will defo go wrong.

Relevant reading:

• all the evaluations linked above
• Some of the other topics of data + evidence
• Freakonomics!
• If you're feeling hardcore, Causal Inference for the Brave and True, my bible.

Non-violent communication is a book written by Marshall Rosenberg. I've never read it, but from my skimming, it is similar to the communication practice taught at Community Mediation Maryland. However, there are some key philsophical differences, and I think it's aimed at different goals. I don't know for sure because I haven't read it yet! Generally it is quite popular among mediators.

Possible method of engagement: a one-time book club meeting to discuss this book, with the expectation that everyone in attendance has read it before attending. If desired, I could bring exercises to practice the skills. Noting that I have not read it before, but I do come at it with an expert eye and some familiarity with the broad topic, this would be like a self-guided skills workshop.

Exercises would be like: simulated role play, where you're given some character background and have to act out a conflict. A chance to practice the language. I don't really know for sure yet what it would look like, having not yet read the book.

Government is facing a problem: AI is making it easier and easier for citizens to get stuff done...and it turns out government was never really designed for it.

It was only a decade ago(ISH) that we all hoped government could be made transparent and responsive through civic technology, hence freedom of information act and sites like What Do They Know.

But today, when anybody can ask ChatGpt to draft and comprehensive request, government departments find themselves sinking under demand, in an AI arms war.

It turns out friction is kind of important! Richard Thaler describes this as "sludge": friction on the system that's seemingly tiny, but enough to affect behaviour.

A world where we all have an AI agent might be the death of friction, and the trend of the last few decades which sees democratic governments being ever more transparent and accessible... But somehow less trusted than ever.

Let's have a bit of a discussion around efforts to make government more transparent, how that didn't seem to be working, and how just maybe we can design some intentional sludge to fix it?

Asymmetric information arises when one party knows more than another. Understanding the implications of this will help you understand everything from why private health insurance is difficult to why socialising with new people is hard.

Asymmetric information occurs in a range of contexts:

• the buyer doesn't know if your car has a hidden fault
• the health insurer doesn't know if you have a hidden condition
• your car insurer doesn't know if you bother locking your car
• your boss doesn't know if you're working or scrolling Tiktok
• you don't know whether or not your new acquaintance is a weirdo

The first three are hidden facts. The second two are hidden choices. Economists study these situations because they can lead to markets unraveling and collapsing. There are also a range of ways to mitigate them including: warranties, references, qualifications, reviews, medical questionnaires, insurance excesses.

For a seminal paper on this issue, see: The Market for "Lemons": Quality Uncertainty and the Market Mechanism (1970)

I define a decision as the process of selecting one out of two or more alternatives.

In many cases, it can be valuable to put some work into generating as many ideas as possible. This is especially important for making decisions amidst disagreement — it takes some real creative work to find an idea that everyone will agree to.

In addition to finding creative ideas to reach agreement, you also simply want ideas that are good! If you are designing a technology, or planning a campaign, you don't just want to settle on the first viable idea that comes along. You want to push yourself and your group to come up with the best possible idea.

Colloquially, the process of coming up with ideas is known as "brainstorming." For design purposes, you must address two key tensions in brainstorming.

On the one hand, you want to maintain as much diversity as possible in your process — the more diverse your ideas are, the more likely you are to find a good one. If all of your ideas are very similar, you're not really exploring the possibility space. Thus, a good design practice might be to keep all of your group members isolated, so they don't start to herd together around the same set of ideas.

On the other hand, people inspire each other — we "riff" on each others ideas — once I hear an idea from you, it sparks an idea from me that I would not have thought about on my own.   So we don't actually want to keep people separate, do we? We want to keep them together, so they can bounce ideas off each other.

This is the tension you must consider in your design process — balancing diversity and creative synergy.

Another technique for generating ideas is crowdsourcing. The idea of crowdsourcing is that we can recruit members outside of a group to contribute ideas, often through a competition where the best idea gets a prize.

The advantage of crowdsourcing is that it puts a lot more people on the problem—and more people means more chance of finding a good idea. A very popular website for crowdsourcing is Kaggle.

One interesting idea I recently read (no link available) about crowdsourcing is that the type of feedback matters — sometimes, you might want to give feedback on the quality of the solutions, and sometimes you might want to give feedback on the way that people are designing the problem.   Each approach has different advantages.

These are a few examples of design principles you might consider, and we can discuss others.

With respect to developing your facilitation practice, I teach a very specific set of techniques for managing idea generation.

Causality is a branch of statistics which economists specialise in. It lets us study all kinds of interesting questions, like the impact of minimum wage on jobs, an extra year of education on income, whether a drug works, smoking bans on smoking rates, pornography on sexual assault.

Let's talk through causal analysis intuitively as well as nerding out on causality theory, e.g. the work of Judea Pearl.

See A Crash Course in Good and Bad Controls (2022), The Book of Why (2019) and Causality (2009).

With the reach of generative AI, the term “slop” has risen in prominence. It represents a rejection of thoughtless content, often perceived to be low in quality or creativity.

Developers have always taken a view about the amount of effort they put in to a task - with a direct effect on the quality of code they deliver. This is a calculation based on deadlines, availability, resources, reputation, and plenty of other factors.

Opinion is strongly divided about AI-assisted coding. Some argue it supports increased productivity. Others cite poor quality outputs and unexpectedly high cost.

What does poor quality code look like?

I have been exploring the outputs from some generative AI coding tools. Some traits are pretty clear:

• Use of obscure and made-up terms

• Code mingled with narrative comments

• Spaghetti code

• Long or difficult to read functions

• Inflexible, hard-coded assumptions

• Dependencies on old libraries, difficult to upgrade

These issues are also referred to as technical debt.

Does it even matter?

Arguably, if code is only ever to be written by one person (or agent), then idiosyncrasies don't matter. Provided the developer knows where everything is, who cares if nobody else can figure it out? This may well be a naive position...

• Do you trust your coding agent to get everything right?
• What happens when you run out of credit?
• Do you want to be able to maintain it yourself?
• Are you sure nobody else will look at the code?

What can we do about it?

Over time, technical debt builds up in every project and the solution is to refactor the code - a sense-making process in which developers rewrite the offending parts. Solo developers and small teams working in isolation are particularly vulnerable to technical debt, but all teams experience it.

Running code reviews and writing code that any person can quickly understand is a good preventative measure. Some examples of established practices:

• Clear and meaningful variable names

• Use of well established design patterns

• Use of existing frameworks, rather than re-inventing the wheel

• DRY (Don’t Repeat Yourself)

• KISS (Keep It Simple Stupid)

• YAGNI (You Ain’t Gonna Need It)

• SOLID (principles for object oriented development)

• Separation of concerns (a precursor to the Single Responsibility Principle)

• Clean Code (and opinionated superset of principles and approaches)

How are coding agents any different?

Coding assistants work alone. They aren't trained to expect code review nor do they have code quality metrics. As they work, they develop personal shorthand and assumptions, which then shapes and styles the code they produce.

I have been collating a set of guidance, skills, and scripts for coding assistants, to raise the quality of their outputs. It's available here: instantiator/dev-qual

For this topic, we’ll discuss factors that contribute to code quality, and how they can be measured and applied - both to AI-assisted and human developed coding. With some metrics in hand, I’ll support you to compose rules for your language or framework of choice, or enhance those already found in dev-qual.

Some time ago, I started building an AI negotiation coach. A negotiation coach is someone who helps a person prep for a negotiation via the sokratic method and, on a limited basis, expert advice. It follows the principle "I am an expert in negotiation, but you are an expertin YOUR negotiation."

I never finished this, as detailed below. I offer this project up to anyone who is interested in building it, or some version of it, or building a completely different thing inspired by it. I can tell you how to do negotiation coaching, and advise you on the lessons I learned so far about building with LLMs.

I abandoned this project after realizing two things:

1. Building things is not a priority for me (The hard part isn't building, it's testing and quality control!)
2. I was not entirely sure how to do negotiation coaching

Since then, two things have happened

1. I have begun doing negotiation coaching, and developed a practice
2. AI has gotten a lot better

The URL below shows the half-baked app. It shows the basic logic and flow, but it uses old models, has no quality control and I honestly don't even remember what process it's attempting to follow.

https://planning.negotiation.solutions/

Software development is necessarily social.

Developers need to communicate well for plenty of reasons:

• to share an understanding of common goals and desirable outcomes

• to share standards for quality

• to share contracts and interfaces for interoperability

• to share patterns for efficiency and readability

• to share learnings to avoid common pitfalls

• to write clear code that’s easy to review

Over the past few decades a variety of ways of working and digital tools have been developed to support this. The agile manifesto and agile software development offers principles and ways of working for teams, and  lean software development goes further, applying lessons from manufacturing to support larger organisations and optimise cross-team communication. Tools like Jira, Trello, Asana, GitHub Projects (and plenty of others) offer easily recognised interfaces to manage this communication.

Open source software is an ecosystem of interoperable tools including operating systems, command line tools, browsers, coding languages and libraries. Many interconnected systems are linked together by APIs running common protocols, with common data standards.

AI assisted coding is becoming prevalent, and may alter the way we work together. Developers can create far more complex projects in much smaller teams, or in isolation. Are shared contracts, open APIs, common data types, and interoperability at risk? What are good outcomes, and how can we incentivise them?

For this topic, we’ll discuss and test principles that could contribute to a manifesto for assisted coding. We’ll be thinking about this, and exploring a variety of ways of working. All ideas welcome!

Much of my professional work has involved working between different professions - in particular I have interfaced between officials (in governments and regulators), academics, and civil society. But also even within government I found it useful to understand differing contexts between departments and (e.g.) analysts vs. policymakers - and develop skills to understand and navigate them.

I can talk to Fellows about practical skills and approaches that can be useful in collaborating across different mindsets and professions. This will include considerations of different incentives and pressures, and methods for learning from engagements, building in particular on thoughts I began in this post:

• www.oliver-marsh.com/academic-non-academic-collaborations-also-charity-donations

This topic builds on Systems Thinking: Understanding Political Dynamics.

Simulation models give us a way to explore how politics systems behave, and what might happen when we change or interfere with them. Examples are how opinions spread through a population, how protests or political movements grow, how inequality accumulates, how institutions respond to pressure, or how different interventions might change the course of a system.

A simulation is a simplified representation of a system. We define the things in it, how they behave, and how they interact, then see what larger patterns emerge. Simulations can be mathematical, computational, physical, or even played out by people.

Traditionally, computer simulation required programming and mathematical expertise. LLMs change that. It is now possible to describe a system in ordinary language and have an LLM help turn it into a working model. That makes simulation modelling much more accessible, but it does not remove the hard part: deciding what to include, what to leave out, what assumptions you are making, and whether the results mean anything.

In this topic, I can help you will build simple models of social and political systems and experiment with them. I can help you choose a system, identify a possible mechanism, turn it into a toy model, use an LLM to implement it, and interpret the results.

The emphasis is on learning how to think with models rather than learning to code.

Core reading

• Marco A. Janssen, Introduction to Agent-Based Modeling (open textbook).
• Ezequiel A. Di Paolo, Jason Noble & Seth Bullock, “Simulation Models as Opaque Thought Experiments” (2000).
• John Bryden, Eric Silverman & Simon T. Powers, “Modelling transitions between egalitarian, dynamic leader and absolutist power structures” (2022).
• Joshua M. Epstein, “Modeling Civil Violence: An Agent-Based Computational Approach” (2002).

No, I don't have the answers of how to completely successfully go against the enormous power of Big Tech companies. But I can talk about the various types of efforts being deployed, some case studies (I list a few below, largely as examples), and where they might work together - including internationally, and by bringing in parties who do not normally focus on political technology.

• Investigative Journalism: From the Facebook Papers to 404 Media
• Political stands: From social media age bans to Palantir.
• Public pressure: From Cambridge Analytica to the Anti Data Centre Movement
• Regulation: From antitrust to tech regulation
• Internal efforts: From Project Maven to Trust & Safety
• Lawsuits: From chatbot harms to social media addiction.
• External Research: From Crowdtangle to the Coalition for Independent Technology Research

I can also talk about some of the disadvantages and discomforts that may be involved, for instance where there is division of what "success" looks like, or trade-offs between (e.g.) completely accurate and truly engaging messaging.

There are numerous opinions and camps around the ethics of various technologies (most strongly and recently, AI), which can create frictions even amongst political technologists who are aligned on other values - let alone people who do not consider themselves political technologists. I can provide advice for Fellows who wish to discuss or navigate polarised views around tech. I will not attempt to convert Fellows to any particular perspective, nor help them to do so with others. I will instead help them consider how they can locate their own work, projects, and views within the existing space; and engage with & prepare for potential criticisms and opposition.

My experience in this topic: Besides my broader work in tech accountability and safety, also at the NGO AlgorithmWatch I was responsible for a lot of work at the intersection of differering views on tech, and a spectrum of organisations who took different approaches and stances. I developed AlgorithmWatch's principles for use of Generative AI, building on surveys which showed diverging views of AI in our organisation. I was also responsible for our research strategy, much of which challenged technologies while also often using (and sometimes encouraged use of) the same technologies.

There is growing appetite for alternative ways of designing social media to give users ways to escape X, Facebook, TikTok and the like - and not just in the political technology community. In some quarters, particularly in Europe, there is growing political attention to this topic with high profile initiatives like rebuild.net (championed by former EU Commissioner for Competition Margrethe Vestager) and #SaveOurSocials. Bluesky in particular is being used and discussed ever more widely.

There are various theories of change for how these platforms could displace existing Big Tech and/or protect democracy - many of which are perhaps over-optimistic and need to be challenged and refined.

I can discuss with Fellows about some of the core innovations and methods underlying alternative social media design - such as interoperability and federation, "public value" AI systems, pro-social design, and middleware.

I can also discuss framing questions around the hoped-for benefits of the design. Who do you hope uses it, and how, in order to achieve your desired change (and is that realistic)? What is your conception of your user groups and their needs, and does that need refining or researching? What existing tech, if any, are you hoping to displace - and if so, how?

Some material for starters:

https://www.betterconflictbulletin.org/

https://www.journalofdemocracy.org/articles/the-future-of-platform-power-making-middleware-work/

https://www.prosocialdesign.org/

https://www.oliver-marsh.com/sideways-looks-35-a-better-internet-part-1-also-hayfever/

https://www.oliver-marsh.com/a-better-internet-part-2-also-mermaids/

The problem of addressing false and misleading content online has a long history, but it became supercharged particularly after 2016. It has spawned much terminology - mis/disinformation, malinformation, FIMI, influence operations - and a range of countermeasures, from fact-checking to media literacy to regulations. New techniques - from fake accounts to deepfakes LLM poisoning - keep emerging, generating new countermeasures but also causing new panics.

The topic has become a major one in political technology discussions, and one which constantly generates new efforts and ideas - many of which do not engage well with existing discussions. There have also been changing views on the nature of the problem itself. Is the risk that minds will be changed in ways that swing elections, or cause mass violence, or even just sow distrust? Or targeted operations to deceive elites or radicalise vunerable individuals? Is the veracity or reliability of information really the best way to address online discourse? Should we be focused more on polarisation and radicalisation than deception - or are these all inextricably related anyway? Or is this problem all overblown anyway, with risks of bad solutions - some of which may threaten freedom of speech?

I can help fellows who wish to engage with "disinformation" navigate this history, and work on an informed basis through core questions: What do I think the problem is? Is my proposed solution meeting the right audiences? Has anyone else tried what I want to do? What might critique or resistance to my solution be?

Some material for starters:

https://www.oliver-marsh.com/sideways-looks--29--ai-pril-fools-also-dating-data/

"Why misinformation must not be ignored": https://pubmed.ncbi.nlm.nih.gov/39666498/

"In the age of AI, keep calm and vote on": https://www.politico.eu/article/artificial-intelligence-elections-ballots-vote-chinsinau-seattle-policymakers-technology/

Economic models are often criticised for relying on absurd assumptions, such as perfect rationality. Despite this criticism, these models persist throughout our politics, policy, and in business.

If modelling is the art of simplifying a complex world through assumptions and abstractions, the question is not if these assumptions are true, but whether they are useful. What makes a good assumption? When does a simplification improve understanding vs worsen it? Let's discuss how economists approach these questions, from traditional modelling methods, to more recent developments in behavioural economics that adapt the assumptions with psychological research.

For further reading, Milton Friedman’s The Methodology of Positive Economics (1953)