PhD: Using AI to Reveal the Computational Principles of Neural Competition

PhD: Using AI to Reveal the Computational Principles of Neural Competition

Can the geometry and strength of AI representations predict which objects win human attention? This project tests directional competition first in controlled visual search, then across successive fixations in natural scenes. It connects computational modelling with cognitive neuroscience; the unresolved question is whether model-derived representational pressure explains attention beyond simpler visual and task factors.

The project record places the doctorate in Birmingham Psychology MIBTP, programme 181D, with Clayton Hickey as primary and Wieske van Zoest as secondary supervisor. It lists October 2027 entry and a 23 November 2026 deadline. Successful international awardees at Birmingham receive full tuition coverage and a UKRI-rate stipend; the MIBTP network caps international studentships at up to 30% of its cohort. The £21,805 figure is the 2026/27 reference, not an announced 2027/28 rate. Discuss the model-to-experiment fit and use the exact 181D application link. Official project; Birmingham MIBTP project and funding; Eligibility framework.

Research Problem and Doctoral Work

The advertised problem is how overlapping distributed population representations compete, and how their structure constrains that competition. It proposes divisive normalisation, combining similarity and relative representational strength into directional representational pressure exerted by one object on another. This is more specific than generic image classification or asking whether two objects are similar. The definition and chosen AI model/layers are not yet published as a fixed implementation.

Initial controlled visual-search experiments combine EEG, eye tracking and behaviour. They test pressure predictions against target enhancement and distractor suppression, N2pc and Pd, gaze timing/direction and performance. Subsequent natural-scene experiments ask which objects attract attention, when they are selected and how multiple competitors shape exploration. Tracking successive fixations connects evolving priorities to model predictions. These are the advertised research stages, not optional generic method training. Complete advertised work.

An evidence-based development would compare the proposed pressure against similarity-only, strength-only and basic visual/task explanations, using held-out objects or scenes to test generalisation. Those are analyst-suggested comparison principles, not a prescribed experimental protocol or promised outcome. A model that predicts EEG or gaze is informative without proving that AI and brain use the same mechanism. Eye movements, neural components and manual performance are related but distinct outcomes; their agreement and disagreement are part of the scientific question.

This field has moved towards naturalistic stimuli, multimodal neural/behavioural data and models with interpretable representational predictions. The three papers below illustrate both progress and unresolved interpretation. A public Birmingham dataset already combines behaviour, image descriptions, eye tracking and MEG with calibration/rejected-trial information. It is a possible preparation/reuse example, not confirmed input for this thesis. Existing multimodal dataset record.

School, Programme and Field Context

The degree is registered at the University of Birmingham, School of Psychology, College of Life and Environmental Sciences, through 181D PhD in School of Psychology MIBTP. The programme directory lists this title; that is dated corroboration beyond the persistent supervisor profile, not a new availability check. 180D belongs to Sport, Exercise and Rehabilitation Sciences. The ordinary Psychology PhD page advertises a three-year September route; the relevant MIBTP route is a four-year October 2027 programme, so its duration/application instructions take priority. Partner-university training does not establish a joint degree. Exact programme mapping; Current theme directory; Ordinary programme comparison.

For institutional scale, the university's Psychology page quotes QS 2027 overall 68th worldwide and 10th UK; these are not psychology subject ranks. More directly relevant, REF 2021 UoA4 Psychology, Psychiatry and Neuroscience reports 54% four-star, 40% three-star, 6% two-star, 55.4 FTE. This is historical evidence for a broad institutional research unit, not a rating of the two supervisors or this project. Subject assessment scope.

MIBTP provides quantitative and cohort training across six universities. The published outline, last revised July 2026 but headed 2025, includes programming/statistics with R, machine learning/modelling, a three-month partner-institution research placement, a three-month non-academic PIPS placement, 12 masterclass days and a technical/scientific visit of one day to three months. These indicate training breadth and travel commitments; confirm the 2027 calendar and thesis integration rather than treating the older outline as fixed dates. The MIBTP FAQ describes fees, stipend, travel/conference and consumables support, and a laptop for programme use; these budgets are distinct from personal income. Published training outline; Programme support and award rules.

Supervisor and Lab Assessment

Clayton Hickey, Associate Professor and BSc Neuroscience Programme Lead, studies attention, learning, reward, motivation and naturalistic vision with behavioural, neural and computational approaches. Contact: c.m.hickey@bham.ac.uk. Wieske van Zoest, Assistant Professor and Director of Postgraduate Programmes, studies the time course of visual selection using eye movements/manual responses and sometimes EEG/MEG. Contact: w.vanzoest@bham.ac.uk. Their complementary interests match the proposed transition from representation to selection and gaze; published titles do not establish spare supervisory capacity. Hickey profile; Van Zoest profile.

The project-specific MIBTP profile promises scheduled supervision at least fortnightly, progress goals and training involving supervisors, experienced postdocs and doctoral researchers. It describes shared task timetables and asks at least two weeks' notice for feedback on up to 5,000 words. These are useful stated expectations, not a final supervision agreement. Ask how the two supervisors split computational and experimental work. Published supervision style.

Recent primary studies

Simkova and colleagues: shared similarity structure

Simkova, Doerig, Hickey and Charest, 20 May 2026, Journal of Vision 26(5):7: 63 participants judged similarity for 100 natural-scene images and 100 matching captions. Visual and language judgments converged, and LLM-embedding-aligned image models better predicted similarity structure than category-trained/AlexNet controls, with relationships to visual fMRI responses. This directly motivates using AI representational geometry. It does not establish complete modality equivalence or show that similarity predicts directional competition; that is the new project's test. The published author-repository PDF was downloaded/read on 21 October; 22 October bibliography checks used institutional/publisher records and indexes after direct PDF access failed. Published representation study

Bibliography: Katerina M. Simkova; Adrien Doerig; Clayton Hickey; Ian Charest. Representations in vision and language converge in a shared, multidimensional space of perceived similarities. Journal of Vision 26(5):7 (2026). DOI: 10.1167/jov.26.5.7.

Hickey and colleagues: EEG variability and function

Hickey et al., 2025, Biological Psychology 195:108987: trial-wise linear mixed modelling links target-elicited NT/N2pc variance with reaction time, and finds sensitivity to distractor identity. The contribution is linking multimodal trial variability to attentional function while distinguishing target and distractor cases. It also leaves a suppression-related interpretation open. This supports careful component modelling rather than labelling every N2pc/Pd change a uniquely identified mechanism; association alone is not causal intervention. Primary N2pc/LMM study

Bibliography: Clayton Hickey; Damiano Grignolio; Vinura Munasinghe; David Acunzo. Using N2pc variability to probe functionality: Linear mixed modelling of trial EEG and behaviour. Biological Psychology 195:108987 (2025). DOI: 10.1016/j.biopsycho.2025.108987. Online 23 January 2025; the institutional record lists publication on 28 January; the issue label is February.

Fakche and colleagues: predictions across fixations

Fakche, Hickey and Jensen, 4 December 2024, Journal of Neuroscience 44(49): e0841242024: simultaneous MEG and eye tracking during free exploration decodes upcoming parafoveal object features/categories at approximately 110/165 ms, and fixated-object features/categories at 100/145 ms. Decoding depends on upcoming saccade goals, consistent with staged processing within an approximately 200 ms intersaccadic interval. This is a bridge to successive-fixation predictions, but an image-array MEG result does not guarantee unrestricted natural-scene or EEG replication. Primary free-viewing study

Bibliography: Camille Fakche; Clayton Hickey; Ole Jensen. Fast Feature- and Category-Related Parafoveal Previewing Support Free Visual Exploration. The Journal of Neuroscience 44(49):e0841242024 (2024). DOI: 10.1523/JNEUROSCI.0841-24.2024.

CHBH's EEG laboratory publishes Biosemi and simultaneous EyeLink capability and analysis using FieldTrip, EEGLab and MNE Python. Access requires training, ethics and project procedures, and technical support is identified. This verifies relevant infrastructure, not reserved equipment time, participant numbers or a dedicated AI compute allocation. Confirm those practical resources and a fallback if simultaneous recordings or recruitment constrain progress. Relevant laboratory and access.

Funding, Application and Living Cost

What is covered, and what remains competitive

The Birmingham MIBTP award terms state that all tuition fees are covered for international MIBTP awardees: UKRI contributes the home-fee portion and Birmingham covers the international difference. Chinese applicants are not excluded by nationality, but overseas academic equivalence, English, visa and award selection still apply. Across the six-university MIBTP cohort, international awards are capped at up to 30%; this network-level allocation is not this project’s quota or an individual’s probability of winning. Awards are based on the candidate; the FAQ warns a selected project can become unavailable and an alternative may be requested. Four-year programme status is established; confirm individual award dates and conditions in the offer. Full Birmingham fee coverage; International award ceiling; Candidate-based selection.

The 2026/27 tax-free stipend reference is £21,805/year. The 2027/28 rate is unpublished in the inspected call; do not freeze the earlier amount for four years. Consumables, conference travel and laptop support are research provision, not extra disposable stipend.

Exact application route and timing

Apply by 23 November 2026 for October 2027. The project record gives no cutoff time or time zone; confirm the closing time in the application portal. Use the project's 181D-specific application link, copy the title and primary supervisor exactly, select the advertised research area and Research Council Funding/BBSRC, answer the research-methods question, upload a CV, and complete the separate ED&I form. A proposal is not required: the mandatory upload field may contain a blank document. The personal statement is optional, with N/A permitted if omitted. Contacting the supervisor is encouraged, not mandatory. English certificates may follow the first application and become an offer condition; English transcripts should show modules, grades and overall result. The FAQ requires at least one academic referee and permits a professional second referee when necessary. Institution-specific instructions; Reference framework.

Applicants may choose one project per programme, up to five programmes, with separate applications; one application naming multiple projects is rejected. This title and the brain–body conscious-awareness project both belong to 181D, so they cannot be submitted as two choices under this rule. Birmingham aims to shortlist by late January 2027, interview in early February 2027, usually Zoom for overseas candidates; the consortium allows late January/early February. The application portal redirects to login and was not inspected behind authentication. Selection limit and interview schedule.

Birmingham housing and upfront cash

Official 2026/27 postgraduate quotes include Maple Bank £107/week, £5,335 contract total, Tennis Court £149/£7,425, and The Spinney £167/£8,332. En-suite examples range £189–222/week; studios include £280–360/week. Utilities, Wi-Fi and contents insurance are included. Contracts vary 42/50/51 weeks, and weekly prices are indicative; the quoted total is not automatically 52 weeks or a guaranteed available room. 2027/28 rents are unverified. Dated postgraduate housing quotes.

Using a 52-week planning illustration, £149/week rent equals approximately £646/month; adding an explicit assumption of £450–650/month for food, transport, phone and other personal expenses gives £1,096–1,296/month. These assumed costs cannot determine savings against an unannounced 2027/28 stipend. A £280/week studio alone is about £1,213/month, before non-rent costs. Use the formal award, real housing quote and relocation/visa/dependant needs; these are scenarios, not measured private-market averages.

The cited government visa guidance lists £558 application fee and student health surcharge £776/year of permission, which can outlast the course. If 4.5 chargeable years apply, the surcharge is £3,492 and total with fee £4,050, before flights/deposit; the programme's approximate £3,600 estimate should not be treated as a personalised quote. The outside-London maintenance benchmark is £1,171/month for up to 9 months, £10,539, not a full 12-month cost estimate. Sponsorship evidence and personal exemptions affect what financial evidence is needed; self-held funds normally require 28 days ending within 31 days of application. Confirm the CAS and current rules at visa time. Current visa fee; Health surcharge rules; Maintenance evidence.

Applicant Fit and Preparation

MIBTP welcomes theoretical and experimental bioscience backgrounds. A first or high 2:1 relevant undergraduate degree meets its stated academic route; low 2:1 plus a master's or at least three years relevant postgraduate experience may qualify; 2:2 or below normally requires exceptional additional experience. Overseas qualifications are assessed for equivalence. A master's is therefore not universally mandatory. The exact project lists its research methods without imposing a separate fixed programming-language or prior-EEG requirement. Academic selection framework.

The China country page's percentage/group tables concern PGT master's admission, so they should not be recast as this 181D doctoral competition's guaranteed thresholds. Likewise, the general English table contains several bands and is labelled for 2026; the exact 181D 2027 score or exemption requires admissions confirmation. University guidance normally requires test certificates no more than two years old at programme start. China qualification context and scope; General English framework.

Useful suggested preparation includes explaining representational similarity versus strength, normalisation and directional prediction; showing competent statistics/programming and reproducible prior analysis; and understanding why gaze, EEG and behaviour require separate model checks. Applicants coming from AI should evidence human-experiment reasoning; those coming from psychology should show quantitative learning readiness. These are research-based suggestions, not extra official prerequisites. The initial contact can include a concise question about the model-to-experiment comparison rather than an unsupported claim that a specific network already explains attention.

Career and transferable skills

Career value centres on reproducible modelling, statistics, human-experiment design and neural/eye-data interpretation. The Graduate Futures career profile identifies academic, contract-research and industry paths, but also competitive academic careers and repeated fixed-term posts. Industry-facing AI or human-behaviour work would require demonstrating software and transferable modelling skills; this topic title alone does not establish an AI-engineering job pipeline. It is a research doctorate, not qualification as a neurologist or clinical neuropsychologist. Career tasks and constraints.

Next Actions and Questions to Verify

Before applying

Choose one 181D project; verify individual degree equivalence and English route, the exact portal deadline clock, and required reference uploads. Ask Hickey whether the intended AI representation/normalisation approach and quantitative training expectations match the proposed background. Follow the exact MIBTP link and complete ED&I.

At interview

Clarify the mathematical pressure definition, model/layer and comparison choices, how controlled-search effects will be tested in natural scenes, and how eye/EEG artefacts and distinct outcomes will be interpreted. Ask about supervisor division/current capacity, participant access, compute and recording support, and a feasible four-year schedule including partner placement and PIPS.

Before accepting

Obtain an award letter confirming full international tuition coverage, 2027/28 stipend, duration and project allocation. Confirm the updated training calendar, visa/CAS costs and sponsorship evidence, housing availability and contract length. Do not budget using a research consumables allowance as personal income.

Evidence and Sources

Project, supervision, programme, funding and application details are limited to records saved in October 2026; they do not verify current vacancy status. The published author-repository PDF of the Simkova representation study was read; subsequent checks were bibliographic, not additional full-text reading. The saved materials do not specify the reading scope for the other papers. Visa and 2026/27 housing costs are dated reference figures, and the login-protected application form’s current fields remain unconfirmed.

Exact project, programme and eligibility

• Exact MIBTP neural competition project and Hickey supervision commitments
• Birmingham 2027 MIBTP exact projects, funding and application rules
• MIBTP 2027 application and international eligibility
• MIBTP programme and funding FAQs
• Published MIBTP training outline
• MIBTP theme directory

School, supervisors, resources and primary studies

• Clayton Hickey official profile
• Wieske van Zoest official profile
• CHBH EEG laboratory infrastructure and access
• Birmingham REF 2021 subject research assessment
• Simkova et al 2026: vision/language similarity representations
• Hickey et al 2025: N2pc variability and linear mixed modelling
• Fakche, Hickey, Jensen 2024: parafoveal preview during free vision
• Psychology research degree: school context and general route
• Fakche research dataset supporting free visual exploration

International preparation, cost and career

• Birmingham China qualifications context
• University general international English framework
• Birmingham postgraduate accommodation 2026/27
• UK Student visa fee reference
• UK current student health surcharge
• UK student maintenance evidence outside London
• Graduate Futures neuroscientist career profile