关于代理生成的符号的涌现与可解释性的博士研究项目

View all current projects

Postgraduate research project

Can machines invent a visual language we can read? Emergence and interpretability of agent-made symbols

Funding

Competition funded

Competition funded

View fees and funding

Type of degree

Doctor of Philosophy

Entry requirements

2:1 honours degree

2:1 honours degree

View full entry requirements

Faculty graduate school

Faculty of Engineering and Physical Sciences

Closing date

30 Jun 2027

On this page

About the project

Potential supervisors

Entry requirements

Fees and funding

How to apply

Contact us

About the project

Potential supervisors

Entry requirements

Fees and funding

How to apply

Contact us

About the project

When artificial agents have to communicate by drawing rather than by exchanging words or tokens, what kind of visual language emerges, and when does it become stable, efficient and understandable to humans? This project studies how populations of agents invent shared visual symbol systems through interaction. You will investigate which properties of the task, the agents and the population determine whether and how such systems arise, and develop the experimental tasks and analysis methods needed to characterise them.

Humans communicate through drawings, diagrams and sketches that are grounded in perception and work across linguistic boundaries. Research on communication between artificial agents, by contrast, has mostly relied on discrete tokens or natural language, and there is no systematic account of how visual communication systems emerge and stabilise. Previous work showed that agents can develop sketch-based communication protocols through interaction using differentiable drawing. This project asks the next question: under what conditions do populations of agents converge on stable, efficient and human-interpretable visual communication?

Unlike tokens, visual messages have no predefined vocabulary: any structure or convention must emerge purely from the pressure to communicate, and many existing measures do not carry over. Characterising what emerges is therefore part of the research. The project will combine interactive, cooperative tasks with new measures of compositionality, abstraction, stability and interpretability, to examine questions such as:

• how do task, bandwidth, architecture and drawing mechanism shape the symbols that emerge?

• do heterogeneous populations converge on a shared visual code, or fragment into local ones?

• how readable are emergent symbols to people, and how do they compare with pictograms or early writing?

• how do design and training choices relate to the properties of the resulting communication?

The emphasis can be adapted to your background and interests, whether computational (learning dynamics), analytical (measures and evaluation) or empirical (human studies). The project spans machine learning, cognitive science, linguistics and human–computer interaction, and informs the design of AI systems that communicate transparently with the people who work alongside them.

The School of Electronics and Computer Science is committed to promoting equality, diversity inclusivity as demonstrated by our Athena SWAN award. We welcome all applicants regardless of their gender, ethnicity, disability, sexual orientation or age, and will give full consideration to applicants seeking flexible working patterns and those who have taken a career break. The University has a generous maternity policy, onsite childcare facilities, and offers a range of benefits to help ensure employees’ well-being and work-life balance. The University of Southampton is committed to sustainability and has been awarded the Platinum EcoAward.

Potential supervisors

Lead supervisor

DM

Dr Daniela Mihai

New Frontiers Fellow

Supervisors

Professor Jonathon Hare

BEng (Hons), PhD, FHEA, SMIEEE

Professor

Research interests

• My main research interests lie in the area of representation learning;

• The long-term goal of my research is to innovate techniques that can allow machines to learn from and understand the information conveyed by data and use that information to fulfil the information needs of humans.

Entry requirements

You must have a UK 2:1 honours degree, or its international equivalent, in one of the following:

• computer science

• engineering

• mathematics

• physics

• cognitive science

• a related field

Essential skills:

• solid programming skills (Python preferred)

Desiderable skills:

• experience in deep learning, reinforcement learning, multi-agent systems, computer vision or experimental design

• interest in language evolution or cognitive science

Fees and funding

We offer a range of funding opportunities for both UK and international students. Horizon Europe fee waivers automatically cover the difference between overseas and UK fees for qualifying students.

Competition-based Presidential Bursaries from the University cover the difference between overseas and UK fees for top-ranked applicants.

Competition-based studentships offered by our schools typically cover UK-level tuition fees and a stipend for living costs for top-ranked applicants.

Funding will be awarded on a rolling basis, so apply early for the best opportunity to be considered.

For more information, please visit our postgraduate research funding pages.

How to apply

Apply now

You need to:

• choose programme type (Research), 2027/28, Faculty of Engineering and Physical Sciences

• select Full time or Part time

• search for programme PhD Computer Science (7089)

• add name of the supervisor in section 2 of the application

Applications should include:

• research proposal

• your CV (resumé)

• 2 academic references

• degree transcripts and certificates to date

• English language qualification (if applicable)

Contact us

Faculty of Engineering and Physical Sciences

For questions about applying, email our Doctoral College (doctoralcollege-admissions@soton.ac.uk).

Project leader

For an initial conversation, email Dr Daniela Mihai (A-D.Mihai@soton.ac.uk).