Skip to content

Notes from 2023 symposium on TCD school of psychology#

Rhodri Cusack - Exploring the foundations of cognition in infants with FMRI and deep learning#

  • MISCONCEPTION: infants brains are like human brains but with bits missing (as in fuctions).
  • They are using neuroimaging and awake fMRI to look at babies/infants while they are doing things. What might be the pre-cursors of functions? Perhaps computational modelling might drive this forward.
  • Particular focus on the ventral visual stream (starting at V1). The fronal cortex is activate quite early when viewing movies (2 months).
  • You can group what stimuli are being seen together in a PCA space and they tend to group pretty reliably into large/small objects for e.g.
  • In deep learning, image classifiers often lean towards the texture of an image as opposed to the shape.
  • IDEA: is the initial blurry vision in infants something that could help training res-net, by initially training on blurry images?
  • In adults, different brain circuits have different timescales - The infants overall have much longer timescales (eeg as well as fMRI).

Lorina Naci - Typical and disrupted brain mechanism for conscious awareness in full-term and preterm infants?#

  • Distinction between capacity in the network for conciousness versus actually the content or being concious of something.
  • Higher-order networks don't really tend to develop in infants.
  • Default mode network, executive control network, dorsal motor network (? I need to double check what these are).
  • The brain functional connectivity tends to follow some of the small-world network properties (statically), and if we look dynamically we get a measure of complexity of the network.

  • What is the ontogony of the brain networks in both adults and infants?

  • Olive Healy - Translating research into practice: An overview of InterAcT (Accomplish & Thrive) an assistive technology software platform for supporting autistic people and individuals with intellectual disability#

  • Creating digital tools to help people with autism and intellectual disabilities.

  • For now, they are looking at working on air travel as a first step.

Kristin Hadfield - Effectiveness of a community-led shared book reading intervention in Syrian refugee children: A randomised controlled trial#

  • They did reading with children, but this intervention was not very effective overall, despite the fact that the children seemed to enjoy it - and thoughts were that it was because the parents were not involved. (this was specifically instructed).

Elva Arulchelvan - "Using a novel tES approach to diminish accelerated long-term forgetting in aMCI"#

  • She is considering the hippocampus and the locus ceruleus as the two main areas of interest.
  • tES is used to apply a small electrical current to the scalp, though a great proportion of the current is lost in the scalp.
  • They are also looking the GON (greater occipital nerve) as a potential target, however this is located deep in the brain.
  • GON projects to LC which then projects to HC, so perhaps this is a way to get the current to the HC (non-invasively).
  • There seems to be memory increases with NITESGON, though it seems that is is low effect size. However, there might be some ideas there for dealing the long term consolidation and memory recall.

Vanessa Teckentrup - "A ubiquitous measure of processing speed using digital questionnaire response times: validity and association with demographic, lifestyle, and health factors"#

  • Looking into processing speed.
  • They use the digital questionnaire response time to measure processing speed, and they found that it was a good measure of processing speed (neureka).
  • Rhodri: Would companies use (misuse) this measure on questionaires to try and measure the "intelligence" of their workers or applicatants?

Graham King - "100 Babies: Lessons from Awake Infant fMRI at 2 months"#

  • The research on this has picked up the last decade or so, but in the last five years or so some groups have really started looking at this in awake infants.

Merve Ataman Devrim - "The Association Between Prematurity and Joint Attention during Parent-Toddler Interaction"#

  • ...

Feng Deng - "Lifestyle activities contribute to cognitive reserve in midlife, independently of education, in cognitively healthy middle-aged individuals at risk for late-life Alzheimer's disease"#

  • ...

Qusai Khraisha - "Parenting in Protracted Refugee Situations: Systematic Review"#

  • ...

Redmond O'Connell - Toward Neurally-Informed models of human decision making#

  • How we make perceptual choices in real-time and the effects of this on cognition.
  • For example, you could ask if a visual grating has a left or right tilt.
  • They are using sequential sampling models to help working with this, the idea of accumulating evidence over time - before making a decision. These models tend to well capture reaction speed and decision making in human data. They also use this to model latent psychological processes of interest and then can be compared.
  • These models are a family that often disagree on the components of the decision making process, which can have a very big impact.
  • Two models that highly disagree in terms of the parameters used can both predict the exact same outcomes and both be just as accurate. So these can lead you to believe very different things depending on what you believe. E.g. you could conclude that we process information more slowly as we age, or that we grow more cautious as we age.
  • Their work at the moment would be on choosing between these kind of models, so that we can more directly measure some things in between to relate them to the model predictions.
  • E.g. combining neural state along with activity in different area that follow the diecision making process, e.g. from visual stimuli to information accumulation, and then from this on tho motor planning, and final the execution of the action.
  • How are expectations related to decision making, which has a paper in prep now.

Paul Liston - “Serious Games for Energy Efficiency: Exploring the Role of Behavioural Biases in Energy Decision Making”#

  • They work in the centre for innovative human systems.
  • They are focusing on making these changes due in terms of encouraging behavioural changes in terms of using energy. A big problem with this is that there is a lack of reflection in your behaviour changes in your bill (it is often delayed and sometimes a small saving).
  • So they are trying to apply gaming elements to this (serious games).
  • The EVIDENT serious game has two factors, one around eduction and the other around repair and replace.

Michael Gormley - “Antecedents of the attitudes of pedestrians, cyclists and drivers towards each other”#

  • They have set up a traffic simulator if anyone is looking to use this.
  • They are looking into how socially different populations interact, e.g. drivers, cyclists, pedestrians. For example, why is there so much hate towards cyclists in the media and what could be done about this.
  • Currently, safety is the biggest barrier to cycling uptake (and the evidence does support this). Motorists also purposely act aggressively towards cyclists.

Clare Kelly - “Rethinking Academia in a Time of Climate Crisis”#

  • Clare is working with Anne Urai - Doughnut academia.
  • Direct lobbying by fossil fuel companies and vested interests are effecting this through the funding of political actors.
  • Perhaps, modern academic life is a huge barrier to having any ability to take action on climate change.
  • IDEA: reframe academia (doughnut ecenomics book published in 2017).
  • IDEA: outdated ecenomic thinking has lead to the problems of climate change. We must reform how we think about ecenomics to enable humanity to live safely and thrive on the planet.
  • How could we rethink our norms to achieve safe academic life?
  • About 25% of a population might be needed for causing a tipping point.

Shane O'Mara - "Do psychologists dream of electric chatbots? The brave new world of competence without comprehension"#

  • Shane thinks that the peer-reviewed research is moving much more slowly than the field is going, so twitter and the likes can actual be a good way to get the latest research.
  • Ethics - biased AI and biased data lead to IRL bias.
  • There is thoughts about the federal trade commission looking into this.
  • Idea of the systems being stochastic parrots (they can repeat what they have heard, but they don't really understand it, and do so with some randomness).
  • Perhaps there will be a decrease in the hype around AI, and a move towards more realistic expectations, and then after that there might be a resurgence in the usage of AI.
  • Only two centres in the world are currently capable of running chatGPT.
  • An AI bot trained as a guard couldn't detect snake style video game cardboard box sneaking (TODO for me, need to look into this one).
  • Med chat GPT is coming in as well apparently. Also is using chat GPT to correct your language cheating?
  • Github co-pilot has just released some data on an A/B tests, and it reduces time to reduce the task by about 50%.
  • Blade runner and do androids dream of electric sheep.
  • Starry AI generates anime style images.
  • Bing actually provides some references for what it is talking about.
  • Diederik Stapel perfomed huge research fraud.
  • Are we reaching the limits of the big applications we can currently use AI for? What might change?
  • These tools will enchance productivity and to initiate, for example, happy to use excelformulabot to create your formulae and avoid having to work with deep excel knowledge.
  • ChatGPT, DeepL, google translate give us great tools to work cross country.
  • Garbage in, garbage out still remains true.
  • Can computers ever generate questions that would be interesting?
  • "I can't describe what it is, but I know it when I see it" - Justice Potter Stewart.

Unkowns (me)#

  • What is brain region SC? (suprachiasmatic nucleus) and how is it involved in the ventral visual stream?
  • What are neonates?
  • What is internal reliability? How does a large PCA factor analysis work?

Questions#

  • Question? How many people are MRI haters (Rhodri joke).
  • How do you measure how much information a baby sees in say 2 years?

  • Is there any difficulty in autistic people interacting with a digital platform?

  • ME: could you develop a very interesting model that actually generates the brain network at each time step of development? So as opposed to comparing to a neural network, you are actually generating the neural network at each time step of development.

  • ME: what is consciousness based on what I have read before, what would seem to be a reasonable definition (is it a Karl Friston style free energy principle)

IDEA#

  • Rhodri, at the moment the companies are taking very simple architectures of models and training them on REALLY large datasets. So they don't really think they need to be convinced by how neuroscience works, because at the moment they seem quite happy with what they are doing -- so maybe in a few years time then maybe we will be in position to influence them.
  • Can look into NITESGON and see if there is anything interesting there.
  • Rhodri's comment kind of lines up with this, biology is messy, but the tech sector does not care about this - which leaves an interesting space!

Unrelated ideas#

  • For my thesis, would it be faster to work on just the open data and try do a really good job there, or to do a split and work a bit on that and a bit on Ham's data?
  • Any poster thoughts and questions will be in my notebook.