EleriThomas
Bio
I'm an independent consultant, advisor, researcher and writer working in strategic design and systems change. I work mostly with UK national and local government, public agencies and non-profits on policy, practice and organisational design that can support better collective decisions, particularly in complex and rapidly changing environments. Since I became self employed about a year ago I've also been writing independently and committing more time to engaging with the broader field and others working in this area.
I'm interested in the conditions under which people, organisations and other complex systems become capable of acting coherently, including how humans (plus machines) co-operate effectively, how institutions and organisations cultivate or undermine collective intelligence, and how psychological, emotional and cultural dynamics shape our capacity to understand and act effectively.
I trained in history, urban design and town planning and have worked across economics, engineering, policy and design, and I'm increasingly informed by contemplative practice, consciousness and natural systems, and what existing models for self organisation in natural systems can teach us about effective collective action.
Before I became self employed I worked for Design Council and the Centre for Cities and within local government, and now I also teach at the University of Westminster.
How others can help me
Continue to challenge my thinking and perspective. I'd be interested in connecting with EA-ers who are working on decision-making, human systems and institutions, research, policy and change.
How I can help others
Probably by informing an understanding of how policy and decisions get made in practice and some of the mechanisms that work or don't work in shifting systems.
Comments3
Surely the most effective way of making change is to influence and in fact directly infiltrate the means of change i.e. through policy and policymaking.
I've been working across research/policy and practice for the last 12 years, and yes it's incredibly difficult to build relationships, trust and traction but I think that's more of a reason to actively engage with it rather than to turn away. I've been particularly interested in the energy in the UK from the "YIMBY" movement, Looking for Growth are a good example of this, and the attitude of "you can just do things" - seeing an increasing number of people directly engage with local decision making, becoming councillors, joining boards to directly influence.
Of course it does critically involve realising that it's not sufficient to be "right", you also need to be able to convince others of that and fundamentally to be able to listen to them - that's not easy but without that there no hope, other than through some kind of autocracy...
The part I'm most curious about here is where you say that "keeping humans in the loop will only degrade decision-making."
I'm fairly new to the AI alignment debates, so I'm not sure how widely accepted this assumption is (and I'd love to be challenged on this) but if it is, I find it quite concerning;
It seems reasonable to design systems that can help humans make better-informed decisions, and to act autonomously where that's appropriate, but that's quite different from concluding that humans should be removed from morally significant decision-making because they make decisions "worse."
I think the distinction we need to make is between cognitive reasoning and moral judgment. I can imagine AI becoming vastly better than humans at reasoning about the consequences of different actions, but that doesn't mean we're deciding that because of AI's advanced cognitive reasoning it should determine the objectives or values being optimised in the first place. It could, of course, help us to identify places where our values or objectives might be out of alignment, but ultimately it's a human decision about whether we agree with that, and those checks and balances should be in there throughout. It's a matter of how we design the interface of AI and human systems as to where humans should be in the loop. That also critically means that AI is then providing sufficient information and evidence back to humans in the process to enable them to update these values based on adequate information.
This kind of all points to the conundrum you raised about it not seeming sufficient to just set values today, or align a model to how we currently make decisions on animal welfare, and assume it will work for the future - I think you already seem pretty clear that it won't be, so then the question is how we DO keep humans in the loop to ensure that.
Hi all, introducing myself here: I'm Eleri (or Elli) Thomas.
I crossed paths with early EA and 80,000 Hours while at Oxford (graduated 2012), though I haven't been active in the community since. Since then I've had a rather varied, interdisciplinary career as an urbanist (trained in history, urban design and town planning, and worked across the built environment professions, economics, infrastructure, smart cities, and policy, design, development and strategy). I've been at think tanks (Centre for Cities), advised government/local authorities/developers (Design Council), held various leadership roles in local government, and co-founded a community-led cohousing company.
I'm now an independent consultant, researcher, facilitator and writer (currently at thinkingthingsthrough.substack.com) focused on strategic design and systems change with a focus on helping enable better collective decisions, especially in complex, fast-changing environments and where there are complex bureaucratic or regulatory contexts or legacies. I've worked with local and central government, a handful of forward-thinking consultancies, and various nonprofits and foundations.
Since going self-employed a year ago, I've had more time to write and engage with the wider field, and I'm really keen to maximise the impact of the insight that I have gained from working on the bridge between research, policy and practice. I'm particularly focussed currently on how humans (and machines) cooperate effectively, how institutions cultivate or undermine collective intelligence, and how psychological, emotional and cultural dynamics shape our capacity to understand and act well.
I'm keen to learn from and see how the EA community is thinking about policy and governance, and on AI development and alignment; here in particular my approach is likely to be more human-centred and design/innovation-oriented than the classic rationalist mindset, so I'm genuinely curious how that friction or complementarity shows up in practice.
In September–October I'll be resident at CEEALAR in Blackpool, so taking the opportunity to re-enter the EA world here. :)