Moral Alignment for LLM Agents
Tennant, E., Hailes, S., Musolesi, M. (2024). "Moral Alignment for LLM Agents." arXiv 2410.01639. https://arxiv.org/abs/2410.01639
CV
Education
- Ph.D in Computer Science, University College London, 2025 (expected)
- M.Phil. in Biological Science (Psychology), University of Cambridge, 2020
- Tools: Python, Twitter API, MongoDB, R, Qualtrics
- All code from my research is available here; journal publication pending
- Training in Machine Learning, Statistics, Python, SQL, R, Unix, High Performance Computing, Psychometrics, Stata, Tableau
- B.Sc. in Psychology and Language Sciences, University College London, 2019
- First Class degree
- Student Achievement Prize for academic achievement in Year 3 (top 4 in class)
Work experience
- Mar 2021-Apr 2022: Quantitative Behavioural (UX) Researcher
- JPMorgan Chase & Co
- Duties: Contributed to the development of a Client Experience Index from behavioural data; analysed behavioral and survey data from external and internal clients across multiple projects (up to 10 in parallel); designed and launched large-scale client survey; conducted extra-curricular research on client-focused digital transformation; Presented results to >5 senior management teams
- Tools: Python, Jupyter Notebook, Qualtrics
- Feb-Nov 2021: Data and Behavioural Scientist
- Rooster Insurance
- Duties: Analysed Marketing and Sales data, built forecasts, performed Cohort Analysis. Created visualisations and reports in Google Data Studio. Contributed to the improvement of the premium Pricing model and the analysis of driving Telematics data.
- Tools: Excel, Google Data Studio, Python, Jupyter Notebook.
- Jun-Nov 2020: Behavioural and AI Scientist
- ProdX.ai
- Duties: Co-built a digital platform that helps teams improve workplace productivity; Independently fully implemented Classifier model using AWS Lambda & DynamoDB; Used behavioural insights to design two platform features, improve people analytics & build valid, ethical and explainable Machine Learning models
- Wrote a blog about using behavioural science for productivity
- Aug-Sep 2018: Research Assistant
- UCL Interaction Centre
- Duties: Helped with research exploring habits and behaviours of crowdworkers from Amazon Mechanical Turk. Analyzed qualitative & quantitative data using Excel & SPSS (N=300);
- Co-authored paper presented at ACM CHI 2019 conference
- Supervisor: Laura Lascau (PhD student) & Prof Anna Cox
- Jan-Mar 2018: Placement Research Student
- Stonecroft Children’s Centre
- Duties: Used standardised tests and designed novel tasks to investigate language development
Skills
- Behavioural Science Skills
- Online survey design, experimental design, observational study design
- Recruiting balanced & (more) representative participant samples online
- Ethics: design of consent & debrief forms; ethical design of online observational studies
- Data Science Skills
- Research: agent-based experiment design; Online, experimental & observational human study design; participant recruitment; ethics; presentation
- Python for data analysis, visualisation & Object-Oriented Programming; Packages: pandas, networkx, NLTK, keras, matplotlib, numpy, scikit-learn
- R (& SPSS) for Statistics: linear & generalised linear model, t-tests, chi-square, Bayes Factor correlations, Factor Analysis, Principal Components Analysis, path modelling with lavaan package
- SQL: basic proficiency (from Cambridge & DataCamp courses)
- Techniques: Machine Learning (esp. Reinforcement Learning & Multi-Agent Reinforcement Learning), NLP, Twitter API, Network Analysis
- Data Visualization: (Python, R, Tableau): area plots, matrix visualization, network plots, histograms, Kernel Density Estimate, violin plots, systematic comparisons of multiple agents’ learning
- Technical Tools: shell scripting, SQL, AWS, MongoDB, IPython, Jupyter, git, Docker, virtual environments, Twitter API
- Hackathons: Cambridge Uni. Data Science Soc. (2019, 3rd place); SheCodes Morgan Stanley (2018, 3rd place)
- Languages
- Russian – native
- English – fluent
- German – advanced
Publications
Dynamics of Moral Behavior in Heterogeneous Populations of Learning Agents
Tennant, E., Hailes, S., Musolesi, M. (2024). "Dynamics of Moral Behavior in Heterogeneous Populations of Learning Agents." The Seventh AAAI/ACM Conference on AI, Ethics & Society (AIES'24).
Learning Machine Morality through Experience and Interaction
Tennant, E., Hailes, S., Musolesi, M. (2024). "Learning Machine Morality through Experience and Interaction " arXiv 2312.01818. https://arxiv.org/abs/2312.01818
Modeling Moral Choices in Social Dilemmas with Multi-Agent Reinforcement Learning
Tennant, E., Hailes, S., Musolesi, M. (2023). "Modeling Moral Choices in Social Dilemmas with Multi-Agent Reinforcement Learning." The 32nd International Joint Conference On Artificial Intelligence (IJCAI'23)
The Grammar of Politics, through the lens of Surveys and Web-based Social Network methods.
Karmannaya, E., & de-Wit, L. (2021, January 11). The Grammar of Politics, through the lens of Surveys and Web-based Social Network methods. PsyArXiv .
Monotasking or Multitasking: Designing Tasks for Crowdworkers’ Preferences.
Lascău, L., Gould, S., Cox, A., Karmannaya, E., Brumby, D.. (2018). "Monotasking or Multitasking: Designing Tasks for Crowdworkers’ Preferences." CHI’19 Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems. ACM, New York.
Talks
Seminar: Social Psychology, Politics and Twitter
Seminar at Darwin College, University of Cambridge, Cambridge, UK
IJCAI2023 Paper Presentation: Modeling Moral Chioices in Social Dilemmas with Multi-Agent Reinforcement Learning
Conference Paper Presentation at Shreaton Grand Macao, Macao, S.A.R. China
Student Conference Presentation: Modeling Moral Chioices in Social Dilemmas with Multi-Agen Reinforcement Learning
Student Conference Presentation at University College Lonon, London, UK
LSE Invited Talk: AI <-> Social Science (NLP, Morality and AI)
Invited Talk for a Student Society at The London School of Economics, London, UK
UCL NeuroAI Conference Lightning Talk: Modeling Moral Chioices in Social Dilemmas with Multi-Agen Reinforcement Learning
Lightning Talk at University College London, London, UK
AI Journal Club Presentation: Transformers
Journal Club Presentation at University College London, London, UK
AI Journal Club Presentation: My PhD Research on Moral Agents
Journal Club Presentation at University College London, London, UK
Teaching
Service and leadership
Positions of Responsibility:
- Treasurer, May Ball Committee @ Darwin College, Cambridge (2019-2020)
- Workshop Organiser – Computational Social Science London (2019-2020)
- Departmental Workshop Organiser (‘Where is Psychology Now?’) @ UCL (2018-2019)
- Designed speaker events and debates bridging research and practice in Psychology & Tech
- Proudest achievement: debate titled ‘Humans vs. Machines in Understanding Language’ between Prof Sophie Scott (Neuroscience Professor @ UCL) and Christos Christodoulopoulos (a researcher from Amazon Alexa)
- Chief Editor & Writer @ UCL Psychology Society Blog, The Bedford Bugle (2016-2019)
- Designed project ‘Crossing Disciplines’ investigating interdisciplinary research at UCL
- Year 1 Student Academic Representative (achieved Advanced Certificate) @ UCL (2016)
Voluntary Work:
- Volunteer @ AAMAS 2023 - The 22nd International Conference on Autonomous Agents and Multiagent Systems (London, 2023)
- Volunteer @ CognitionX: Festival of AI and Emerging Technology (London, 2019)
- Behavioural Science Volunteer @ Climate Change Communication Workshop (Cambridge, 2019)
- Impact Research Intern @ Greater Change (remote, 2019)
- Greater Change addresses homelessness by enabling personalised, cashless donations
- Wrote an Impact Report evaluating the charity’s 2-year impact on the lives of clients (for example, integrating the charity’s data with statistics provided by the local council to estimate financial impact on the council’s budget)
- Senior Fellow/Planning Team Leader @ Movement for Hope (London, 2016-2018)
- Movement for Hope creates art-science projects to support research in neuroscience
- Co-created events #Rewired2017 (see Lancet review), #Rewired2018 and a student-researcher video interview project
- Participated in Patient & Public Involvement workshop
- Led the 2018 Planning Team
- Sat on judging panel of final year Medical Physics & Engineering Prototype projects