Reflections & learnings— 2nd year at the Alan Turing Institute

Eirini Malliaraki
12 min readDec 21, 2020


It has been an extremely challenging year — and I’m incredibly thankful for being healthy and having a job. This is a personal blog post, and I will reflect on some of my learnings, contributions, and recurring thoughts during my second year at the Turing. I broke down what I’ve been working on into four areas: exploration, mission orientated R&I, infrastructure and policy. I still have a cross-cutting role, working across and within teams at the early stages of shaping new projects but this year I had the opportunity to dive deeper into some. It’s been inspiring to see what everyone has pulled together and the organisation’s capacity to collaborate in new ways. I’m *really* grateful to every colleague I’ve worked with this year for their resilience, good vibes and continuing support.


Climate Resilience & Security

I spent the first ~ five months of 2020 working on a project on climate security. Climate change is expected to have severe implications for human and national security. Increases in extreme weather events and long-term climate changes can challenge water and food security and exacerbate migration, instability and conflicts that will cost millions of lives and trillions of pounds. Also, transitioning to renewable energy will influence countries’ power balance, reconfigure trade flows, and create new interdependencies around renewables and commodities.

These systemic risks exhibit complex relationships that extend in space and time via flows of materials, people’s movement, and trade linkages. Understanding and governing ‘Anthropocene’ risks will require new and dynamic strategic foresight capabilities and tools that help policymakers capture the feedback loops between social and natural systems that result in resilience changes and identify where resilience needs to be built in and/or safeguarded.

This map illustrates the connected nature of climate security risks which include physical risks to resources and infrastructure, risks to human mobility and social cohesion and transition risks from climate change adaptation and mitigation efforts.

I coordinated a scoping study to assess the feasibility and resources needed to develop a new generation of models and early warning systems that forecast climate risks for the next five years. Our study comprises a rapid evidence assessment of literature, expert consultation, stakeholder engagement and final synthesis. We flag opportunities for data science and AI to help a) determine where the accumulation of climatic stresses is interacting with human populations b) issue early warnings of potential climate security tipping points which can inform proactive resilience planning responses, c) develop new causal models that link climate and socio-economic processes and define the boundaries for safe and just ‘operating spaces’.

This project opened my eyes to climate security futures and the nature of systemic risks. I found that transition risks from climate change mitigation and adaptation efforts are underappreciated. For example, adaptation efforts can aggravate existing inequalities or grievances over resources. On the mitigation side, the ongoing net-zero transition involves a much deeper transformation of the world’s energy systems with major social, economic and political implications. For instance, countries that are highly dependent on fossil fuel rents could lose 12 trillion US dollars in stranded assets and become exposed to transition risk (which can trickle down international financial markets). The project also showed me the limitations of data-driven tools that do not account for more disruptive futures, e.g. if geoengineering will be a significant security risk or if solar radiation management will be weaponised.

These new kinds of planetary-scale risks bring the interconnectedness of our systems and the nation-state’s limitations to the forefront. Existing conceptualisations of globally connected systemic risk poorly capture human-environment interactions that exhibit complex, cross-scale relationships. I think it’s essential for policy and decision-makers to navigate this new epoch with novel tools and fit for purpose governance structures that acknowledge the realities of the Anthropocene.

Modern Slavery Data Policy Innovation

Another project I am involved in is about tackling modern slavery. It is led by Anjali Mazumder and is in collaboration with the Open Data Institute and Oxford’s Bonavero Institute. (It’s crazy how many people today are still in some form of slavery — the International Labour Organization estimates that 40 million people are enslaved, including forced labour, exploitation in the private sector such as domestic work, construction or agriculture, forced sexual exploitation and forced marriage).

Tackling modern slavery requires cooperation amongst government agencies, non-profit and private/for-profit institutions to effectively share data, tools, and resources to deliver services and protect communities and institutions. The data landscape is complex and relevant data is sparse and patchy as often no one organisation follows the path of an individual: Vulnerable and exploited individuals have contact with a wide range of public (e.g. health, law enforcement, border authority), non-profit (e.g. helpline, support services) and private sector (e.g. financial institutions) services, and many cross borders.

In this project, we aim to scope and identify the data and algorithmic trust, security and privacy requirements for policy innovation and infrastructure needed to tackle modern slavery. We will also determine the need for privacy-enhancing technologies and data institutions that enable more responsible data sharing. I was involved in setting up the project plan and timeline, bringing our collaborators together, and researching and mapping the roles and data flows between various data stakeholders, e.g. beneficiaries, contributors, creators, stewards, etc. (learning loads from the insightful work of the ODI). The project is ongoing, and we will publish our findings in spring 2021.

Mission-oriented R&I

Turing AI Scientist Grand Challenge

This is a new fascinating grand challenge project that will explore what is required to develop AI systems capable of making Nobel quality scientific discoveries by 2050. Starting in January 2021, the team will create a study that includes a multi-year roadmap with technical quests and intermediate milestones for scientific discovery in biomedicine, environmental sciences, and materials sciences. I was involved in setting up the project with the Principal Investigators and my colleagues Hushpreet and Will. This includes establishing its identity, participating in the hiring panel for the postdoc, writing and shaping the report’s outline, raising awareness across the Turing and beyond and delivering strategic community engagement activities with other research councils. It is very timely given Deepmind’s recent success in protein folding. The project also raises new and important questions about the nature of scientific discovery, the automation of science, the servitisation of AI and the future of scientific skills. It also inspired me to write about Architecting Moonshots. We also made a map of the landscape for AI in the Sciences in the UK (you can cluster by topic and university.

Climate action

We now have several research projects exploring how machine learning and data science can be part of the solution to climate change. These cover mitigation measures that reduce greenhouse gas emissions; adaptation, i.e. improving our capacity to prepare, plan for, recover and adapt to weather-related disasters, extreme events, and natural hazards; and building a digital picture of our natural environment that allows for better monitoring of the impacts of climate change in agriculture, biodiversity, oceans, land, water, and the cryosphere. Building on work from last year and under the visionary leadership of Environmental Science Theme Lead Scott Hosking, I’m working with various teams to ensure that this is joint up and that our interventions talk to each other. Our efforts are now displayed on our climate action page, and if you want to be involved, you can join our new Environment and Sustainability Interest Group.

Inspired by our work at the Turing, I wrote a few words on AI and climate action for 𝘉𝘳𝘢𝘯𝘤𝘩, a new magazine about sustainable and just internet futures, supported by EIT Climate KIC, Mozilla Foundation, and the Green Web Foundation. I also contributed to a new white paper by WWF and the World Bank on ‘Spatial Finance: Challenges and Opportunities’. In the report, we outline a new taxonomy for spatial finance and show how different data can be combined to complement existing ESG methods. (Environmental, Social, and Corporate Governance are three key factors in measuring the sustainability and societal impact of an investment in a company. If financial markets are to realign towards real climate action, the financial sector needs to assess commercial actors more accurately on their climate and environmental performance.)


Pangeo for Environmental Science

I’m quite excited about our latest work on Pangeo, and I see it as a horizontal enabler for our climate action work. Pangeo is a framework for big data geoscience on the Cloud and high-performance computing that uses open-source components from the Python ecosystem. It allows for interactive and scalable computing on large gridded datasets used by ocean, atmosphere, land and climate scientists. This project is supported by Microsoft AI for Earth. It is part of the new Turing- Met Office collaboration, which will drive research at the interface of data science, AI and climate science (I’m also sitting at the Turing & Met Office Collaboration Steering Committee). I was involved early on in shaping the scope with colleagues at the Research Software engineering group. Now I’m working with the team to ensure our work is appropriately embedded in our broader community.

This project opened my eyes to the infrastructural challenges of data scientific work that hinders environmental researchers (in particular) as well as the intricacies of community-led software projects. It made clear the need to support communities that build the backbone of research software and develop end-to-end data pipelines with much-needed data cataloguing, discovery and provenance features. This work is aligned with our Tools Practices and Systems programme led by the exceptional leader Kirstie Whittaker with whom I’ve had the opportunity to bounce off ideas about strategy and community engagement.


Public Policy Consultations

This has not traditionally been part of my role, but I was drawn to it this year quite a few times. I helped coordinate and write parts of the Turing’s response to public consultations by the Government. These consultations include the UK’s industrial strategy, the development of the British Arpa, and the National Data Strategy by DCMS. For the latter, I coordinated and synthesised our response that combines the perspectives of various Turing staff and researchers affiliated with the Turing. I worked with Amit Mulji and Christine Foster, and we tried a variety of approaches to harness the Turing’s collective brain — sort of A/B testing the consultation process. There is a lot of tacit knowledge in network organisations like the Turing — not just in our academic network, but also in the business team, the legal team, the research software engineering team, the programme management team etc. and I’d like to encourage more diverse crowds to participate in shaping policy.

Covid and AI conference and report

This activity is part of the Turing’s effort to document and evaluate the UK AI & data science communities’ response to the pandemic; highlight lessons learned and make recommendations going forward, including future research, and best practice. It comprises a conference that explores the lessons learned during the COVID-19 pandemic and a series of workshops exploring the role of AI and data science during the pandemic to understand how the research community can be better positioned to respond to major crises in the future. I worked with Anjali Mazumder and Inken von Borzyskowski, for the conference, and I did a rapid data analysis of covid related AI publications since the beginning of the outbreak using Microsoft’s Academic Graph. Preliminary findings were presented at the conference. I was also involved in a selection panel for the workshop participants. I’m looking forward to the conclusions and policy recommendations that the team will publish in early 2021.

Side projects

Covid creatives toolkit

This is a crowdsourced mutual aid compilation of free & open-source resources that aims to support creative practitioners going digital in the pandemic. The toolkit began as an open call initiated by Kat Braybrooke in the early days of social distancing in 2020. Kat invited me to join other curators from across the arts, technology, community, academia, and gig work. Together, we built an open archive of mutual aid resources for creative practitioners who found themselves needing to migrate their practice onto digital spaces. The kit’s 7 sections reflected on the virus’s going-digital requirements, from co-creation tools to collective action platforms, self-care apps and other much-needed offerings that promoted well-being.

Mapping COVID-19 Conspiracy Tribes Across Instagram, TikTok, Telegram and YouTube

This is a project I took part in during the Digital Methods Summer school at the University of Amsterdam in June 2020. The project investigates how several popular conspiracy theories (including 5G, Bill Gates, QAnon, flat earth, and the deep state) spread across social media. We analysed data from four platforms: Instagram, TikTok, Telegram, and YouTube. We found that engagement with these conspiracy theories remains relatively low when the virus was still contained to China but started to increase in March and peak in April/May when COVID-19 became a global health crisis. Each platform analysis yielded specific findings. In my group, we looked at Instagram and found that in January and February, COVID-19-related hashtags were present in much lower volume than in March and April when they also started to act as a linchpin bringing disparate conspiracy theories together. A full analysis can be found here.

I enjoyed this project because it allowed me to get hands-on with data and work with journalists and cultural theorists. I got exposed to the high weirdness of online conspiracies, new obfuscation tactics, absurd memes, the belief systems of these communities, and the dynamics of cross-platform conspiracy spreading etc. These conspiracy narratives pose obvious new challenges for sensemaking in society and social media platforms grappling with the rapid and large scale dissemination of dis- and misinformation in their networks. Several efforts are trying to detect and alleviate infodemics, but fewer are trying to understand these communities’ underlying knowledge culture, tactics, and motives.

1k most-liked Instagram photos associated with conspiratorial content from Feb-June. ~50% of them are screenshots from Twitter.

What’s on my mind lately

Well, the state of our world in crisis and the topics I’ve been exploring above, i.e. how can we add self-awareness to Gaia’s self-regulation, live more sustainably and make better sense of our polarised world. There are a couple of crosscutting themes that I keep coming back to lately:

Resilience and institutional agility

Through this year’s projects & engagements, I’ve noticed a disconnection between the timeframes that policy operates in and those of our social and environmental systems. I also see a disconnection between research policy and funding and the world of catastrophic risks such as climate change or advanced AI. For example, operational and programmatic decisions at the national and international level take place on a 3–5 year horizon. In the context of environmental change, the relevant horizon is decades (this is the difference between weather and climate).

Our institutions require much more agile and longterm thinking, supported by appropriate tools and practices. To borrow the words of Indy Johar, we need to move away from “solutioneering” and towards building the capacity of a system to respond and be aware of its interdependencies. This need became evident during this pandemic. The next decade will be volatile and will bring more shocks and crises. I’m curious to understand the types of structures that support organisational longtermism and agility (such as reporting lines, knowledge management systems, oversight processes, team organisation, etc.). If we understand the factors that contribute to institutions’ ability to mitigate existential risk and deliver long-term strategic goals, we may inform future regulatory infrastructure and policies that encourage these attitudes.

Collective wisdom

In the past year, we saw unprecedented collaboration and sharing of knowledge and resources — in academic communities, the healthcare system, in the self-organised mutual aid groups etc. I keep coming back to T S Elliot’s quote “Where is the knowledge we have lost in information? Where is the wisdom we have lost in knowledge?”. I think we should be talking more about the state of our collective wisdom, i.e. our ability to discern the interconnections, interdependencies and resonances of our hyperconnected lives.

Why now? More people are online and are creating and connecting digitally more than ever before. This exponential increase in the quantity of text/media combined with a chronoscopic timeframe poses new information overload and sensemaking challenges. We’ve also had the chance to collectively pause and reflect on things like our health, what is valuable, what is essential etc. Thus, there is potentially existential significance of that insight for the individual that calls for a radical reframing of knowledge and our relationship to it (I talk about building my second brain here). Lastly, there is a pressing need to cultivate more fruitful dialogue between emerging and increasingly polarised perspectives. The more complex the problems are, the more likely it won’t be connected agents who will solve them, but communities of interconnected agents so I think we need to cultivate their:

  • systemic awareness (grow the resources and activities that make us aware of the social, natural and technological systems we are part of so we can adjust our actions accordingly);
  • systemic health (use new technologies, systems and movements to build healthier communities and ecosystems — I’ve written about One health approaches);
  • Systemic learning (better support communities, movements, and societies to learn, grow in wisdom, and act coherently while ameliorating collective stupidities like groupthink and conspiracies.)

Lots to think about and take forward in 2021 — in the meantime, let’s be patient and hold each other up.

* if any of these thoughts resonate with you, please reach out :)