1 of 43

Barriers to learning - a science of learning is incomplete without understanding variability

Duncan Astle

2 of 43

Key messages for this talk

  • Expect variability in children’s cognitive skills… and plenty of it!
  • This variability is an inevitable (and very natural!) consequence of the mechanisms of brain development
  • This variability in cognitive skills cannot be “fixed”
  • Supports and modifications crucial to make mainstream classrooms accessible

3 of 43

Genes

Cells

Networks

Systems

Behaviour

Developmental Neuroinformatics

Cognitive development varies widely across children. Worldwide ~15% would meet criteria for a recognised neurodevelopmental condition.

These differences are often life-long and can act as barriers to education, wellbeing, healthcare and employment.

How much variability should I expect?

Where does this variability come from?

What does this mean for education?

4 of 43

How much variability should I expect?

A large cohort (N=1000) enriched with children at ‘neurodevelopmental risk’

Most children come from 15 different referral routes within Children’s Services, including CAMHS, Paediatrics, Educational Psychology, Specialist Teachers

Children can come with single, multiple or no formal diagnostic label

Genes

Brain Structure

Brain Function

Cognition

Learning

Behaviour

Mental Health

All in mainstream education, mostly aged 7-11, 67% boys

A rich multilevel dataset, fuelling >75 projects in Cambridge so far

5 year longitudinal data collection almost complete

Now a managed access database

5 of 43

Unsupervised machine learning: Simple artificial neural networks

v1

v2

v3

w1

w2

w3

Kohonen, T. (1982). Biological cybernetics

Dittenbach, M., Merkl, D., & Rauber, A. (2000). IEEE

Size

Acidity

Seeds?

Size

Acidity

Seeds?

6 of 43

SIZE

ACIDITY

SEEDS

Kohonen, T. (1982). Biological cybernetics

Dittenbach, M., Merkl, D., & Rauber, A. (2000). IEEE

Unsupervised machine learning: Simple artificial neural networks

Why use this machine learning approach?

  • You might learn something
  • Creates a model that can be used to make predictions

7 of 43

‘Berries’

Why use this machine learning approach?

  • You might learn something
  • Creates a model that can be used to make predictions

Unsupervised machine learning: Simple artificial neural networks

Kohonen, T. (1982). Biological cybernetics

Dittenbach, M., Merkl, D., & Rauber, A. (2000). IEEE

8 of 43

Memory

Vocabulary

Listening skills

Attention

CALM Referrals:

Cognitive Battery

ADHD?

Dyslexia?

Autism?

Other stuff?

9 of 43

Siugzdaite et al. (2020). Current Biology

Unsupervised machine learning: Simple artificial neural networks

Cognition

Around 2-3 years ahead of what you would expect for their age

Around 2-3 years behind what you would expect for their age

10 of 43

What predicts your location in cognitive space?

Seeing the Speech & Language Therapist

Dyslexia diagnoses

ADHD diagnoses

Diagnosis is not a good predictor of a child’s cognitive profile

Siugzdaite et al. (2020). Current Biology

Autism diagnoses

11 of 43

What predicts your location in cognitive space?

Siugzdaite et al. (2020). Current Biology

Quality of Prediction

Learning

12 of 43

What predicts your location in cognitive space?

Siugzdaite et al. (2020). Current Biology

Quality of Prediction

13 of 43

What predicts your location in cognitive space?

Siugzdaite et al. (2020). Current Biology

‘Modularity’

‘Integration’

‘Centrality’

Cognitive performance

More ‘small world’

More random

Quality of Prediction

14 of 43

What have we learnt so far?

Cognitive profiles of children in mainstream school differ widely

These differences do not align well with formal diagnostic labels

But they are related to individual differences in attainment

How children’s brains are organised is predictive of these differences in cognition

ADHD diagnoses

Quality of Prediction

15 of 43

How much variability should I expect?

Developmental Neuroinformatics

Where does this variability come from?

16 of 43

Using generative networks to simulate brain development

Akarca et al. (2021). Nature Communications

“What I cannot build, I do not understand” Richard Feynman

17 of 43

Complexity emerges from simplicity

Kaiser & Hilgetag (2004) Neurocomputing

Vértes et al. (2012). PNAS

Betzel et al. (2016) Neuroimage

18 of 43

Complexity emerges from simplicity

Akarca et al. (2021). Nature Communications

Rapid updating

of wiring probabilities

19 of 43

Simulating network formation

Akarca et al. (2021). Nature Communications

Evaluate with an

energy equation

Evaluate energy

for different parameter

combinations

Evaluate energy for different K terms

Different K terms

Dan Akarca

20 of 43

Tiny amounts of variability generates diverse trajectories

Akarca et al. (2021). Nature Communications

Simulating network formation

The low energy simulations come from very specific parameter combinations

21 of 43

So where does network diversity come from?

Akarca et al. (2021). Nature Communications

22 of 43

It’s the trade-off that produces diversity

Why?

Akarca et al. (2021). Nature Communications

23 of 43

What shapes the trade-off?

Monaghan et al. (2023). In review

Alicja

Monaghan

Three sets of genes

  1. PGS gene
  2. ‘Cost-associated’ genes
  3. ‘Value-associated’ genes

PGS genes

‘Costs’ genes

‘Values’ genes

Genes associated with cognitive ability also associated with the values part of the equation

Sofia Carozza

Carozza et al. (2023). Developmental Psychobiology

Early environments causally shape the parameters

24 of 43

What have we learnt so far?

We can simulate the formation of children’s brain networks by trading off the ‘costs’ of forming a connection with the ‘value’ that connection brings

For each child, this trade-off is slightly different

These small differences generate diverse trajectories of brain development

In turn these differences are likely shaped by both early environmental experiences and genetics

PGS genes

‘Costs’ genes

‘Values’ genes

25 of 43

Developmental Neuroinformatics

How much variability should I expect?

Where does this variability come from?

What does this mean for education?

26 of 43

So we boost cognition… right?

A long history of training interventions designed to boost cognition directly

These interventions can alter neural networks in children’s brains

But they have no wider benefits

27 of 43

The Diverse Trajectories to Good Developmental Outcomes Workshop (December 2022)

If we cannot change cognition, then how do we make education inclusive?

Four targeted focus groups

100 expert delegates from a diverse range of fields and lived experiences

Four evidence-based symposia followed by open discussions

Co-create resources to guide schools about inclusive practices

28 of 43

Background: inclusion in the UK

  • Inclusion for children is mandated practice in the United Kingdom (in England this is framed around SEND).
  • In 2022, there were an estimated 1.4 million state students with SEND in England.
  • In England the diagnostic labels that often gate access to support give only a limited and static snapshot of a student’s needs.
  • Social inequalities in access to support interact with learning barriers, compounding the exclusion problem. Boys and ethnic minorities are most likely to be excluded.

“The system for supporting pupils with SEND is not, on current trends, financially sustainable” National Audit Office (2019, p.11)

“even significant funding increases might make little difference to children and young people with SEND” House of Commons, Education Committee (2019, p.3)

29 of 43

Key Message #1: Children with SEND are disproportionately likely to face bullying and exclusion

To target this, we can:

  • Normalise differences in the classroom & celebrate diversity
  • Equip the workforce for inclusive practices
  • Eliminate unfair school incentives (such as attendance prizes)
  • Offer solutions and accommodations, rather than punishing students for having support needs

What did we learn at the workshop?

30 of 43

Learning About Neurodiversity at School (LEANS)

LEANS is a new resource pack to introduce neurodiversity concepts to mainstream primary classes in an accessible and engaging way.

Download LEANS for free and find out more!

Overview talk this Thursday June 22nd!

UCD Neurodiversity Masterclass Seminar

Online, free (17:00-18:00)

https://www.eventbrite.ie/e/neurodiversity-masterclass-series-tickets-140051587067

Or visit the LEANS event page for booking link

31 of 43

Key Message #2: Universal approaches to teaching can facilitate learning for all students

To enable all learners to flourish, the UK’s educational system needs to move from inclusion on demand to inclusion by design.

Universal Design for Learning (UDL) systems don’t cater to a minority of students - they consider all students’ needs and create open and accessible systems for everyone, neurotypical or neurodivergent, diagnosed or undiagnosed.

Provide multiple means of Engaging with learning

Provide information in multiple Representational Formats

Provide multiple means of Demonstrating Knowledge

What did we learn at the workshop?

32 of 43

Key Message #2: Universal approaches to teaching can facilitate learning for all students

What did we learn at the workshop?

33 of 43

  • Recognise exclusion

  • Learn from diversity

  • Solve for one, extend to many

Key Message #2: Universal approaches to teaching can facilitate learning for all students

What did we learn at the workshop?

34 of 43

Key Message #3: Schools should aim to co-produce inclusive policies with parents, staff and students with lived experiences of neurodivergence

What did we learn at the workshop?

35 of 43

Key Message #4: A neurodiversity-oriented approach can promote a sense of belongingness at school

  • Evaluations of high exclusion rates among neurodivergent children have shown that problems often begin when a child’s needs are not being met.
  • Neurodiversity-based approaches can help shift attitudes across a school. Evidence from the LEANS curriculum suggests that informing students about neurodiversity can change their attitudes about differences among their peers.
  • By investigating why students feel excluded and co-producing school policies with students and parents, educators can learn about their students’ needs and foster a greater sense of belonging.

What did we learn at the workshop?

36 of 43

Belonging at School

Coming in September 2023!

New FREE resources for educators about neurodevelopmental differences, school inclusion, and co-designing inclusive policies.

Join our mailing list so you won’t miss them!

Scan QR or mail inclusive-education-subscribe@lists.mrc-cbu.cam.ac.uk

37 of 43

What did we learn so far?

Attempts to “fix” cognitive differences have so far had poor results

With around 1.4m children in England having additional needs, this requires that we make schools more inclusive places

We can do this by shifting attitudes about differences

Moving to inclusion by design, rather than inclusion on demand

Plans need to be local and co-produced with children and parents within schools

The end goal is to make schools a place where children feel they belong, regardless of their differences

38 of 43

We should expect a lot of variability in cognition

This is a natural consequence of the mechanics of brain development

So we need to make sure that schools are inclusive places that can accommodate these differences

39 of 43

Find out more at: www.astlelab.com

Twitter: @duncansastle

Current Funders:

40 of 43

4D Research Group

Susan Gathercole

Joni Holmes

Tom Manly

Kate Baker

Rogier Kievit

Joe Bathelt

Giacomo Bignardi

Sarah Bishop

Erica Bottacin

Lara Bridge

Annie Bryant

Sally Butterfield

Elizabeth Byrne

Gemma Crickmore

Fánchea Daly

Edwin Dalmaijer

Tina Emery

Laura Forde

Delia Fuhrmann

Andrew Gadie

Sara Gharooni

Jacalyn Guy

Erin Hawkins

Agniezska Jaroslawska

Amy Johnson

Jonathan Jones

Silvana Mareva

Elise Ng-Cordell

Sinead O’Brien

Cliodhna O’Leary

Joseph Rennie

Ivan Simpson-Kent

Roma Siugzdaite

Tess Smith

Stepheni Uh

Francesca Woolgar

Mengya Zhang

Find out more at: www.astlelab.com

Twitter: @duncansastle

The CALM Team:

GNMs

Petra Vértes

Ed Bullmore

Collaborators:

MEAs

Stephen Eglen

Ole Paulsen

Manuel Schroeter

Alex Dunn

Biophysical RNNs

Jascha Achterberg

John Duncan

Matt Botvinick

Child Development

Gaia Scerif

Kia Nobre

Sarah-Jayne Blakemore

Tamsin Ford

Derek Jones

Mark Johnson

Sue Gathercole

Joni Holmes

Maros Rovny

PhD student

Chris Iordanov

Placement Student

41 of 43

Belonging at School

Scan QR or mail inclusive-education-subscribe@lists.mrc-cbu.cam.ac.uk

Resources for you

42 of 43

43 of 43