Senior Research Associate in Machine Learning for Spontaneous Inner Speech Detection

Inner speech – talking to yourself in your mind – appears fundamental to human consciousness, thinking, and self-reflection. Yet we have no reliable way to objectively detect or measure it as it happens spontaneously in everyday life. This project tackles one of cognitive neuroscience’s most challenging problems: detecting fleeting, spontaneous inner speech from the “haystack” of ongoing brain activity.

The Challenge

Can we objectively detect inner speech – the voice in your head – from brain signals? This project tackles one of cognitive neuroscience's hardest problems: identifying spontaneous inner speech from noisy EEG data without precise temporal labels.

Traditional classification has failed because spontaneous inner speech is sparse and co-occurs with other brain activities. We need novel ML approaches suited to weakly-supervised settings, transfer learning, or contrastive methods to detect these fleeting cognitive events.

Your role

Working with Dr Bo Yao (Lancaster) and Professor Xin Yao (Lingnan University, Hong Kong), you'll develop and validate a novel ML approach for inner speech detection from high-density EEG data.

Deliverables:

• Working implementation of one novel detection approach
• Systematic validation against baseline methods
• Lead manuscript for publication
• Documentation of what works, what doesn't, and why

This is fast-paced, requiring rapid prototyping with access to Lancaster's high-performance computing facilities.

Essential Requirements

• PhD in Machine Learning, Computer Science, Computational Neuroscience, or related field
• Demonstrable experience developing deep learning models for time-series/sequential data (EEG, biosignals, audio, sensor data, or similar)
• Strong Python skills with PyTorch or TensorFlow
• Proven ability to work independently on complex problems
• Excellent communication skills for interdisciplinary collaboration
• Right to work in UK for project duration

Desirable

• Experience with ML for unlabelled/sparsely-labelled sequential data (self-supervised learning, anomaly detection, domain adaptation)
• Model interpretability techniques (attention mechanisms, saliency mapping)
• Prior work with neuroimaging data or biosignals
• First-author publications in ML or computational neuroscience

Why This Role?

• Intellectual freedom – real ownership of methods
• Cross-disciplinary experience – apply ML expertise to neuroscience
• Fast-track impact – see your methods in use within months
• Flexibility – 0.8 FTE for work-life balance
• Strong mentorship – collaboration with experts in neuroscience and AI
• Career development – potential first-author publication(s) at the intersection of AI and consciousness science

Benefits

• 25 days annual leave (pro-rata) plus closure days and bank holidays
• Pension scheme and flexible benefits
• Athena Swan Silver Award department
• Flexible working arrangements

To Apply

Submit via Lancaster University Jobs Portal:

1. CV (standard academic format)
2. Cover letter (max 2 pages) addressing:

• • Your most relevant ML project with time-series/noisy/weakly-supervised data (your role, methods, outcomes)
  • One potential approach for detecting sparse, unlabelled events in noisy multivariate time-series
  • Why this project appeals to you now

Code sample (optional): GitHub repo/notebook demonstrating your implementation style. Informal enquiries encouraged: Dr Bo Yao, b.yao1@lancaster.ac.uk

Type / Role:

Subject Area(s):

Location(s):