Alex Pike
LightNews LightNews
banner
alexandrapike.bsky.social
Alex Pike
@alexandrapike.bsky.social
1.9K followers 1K following 210 posts
Lecturer in Mental Health @YorkPsychology || eating disorders, anxiety disorders, cognition/computation || she/her

Group page: https://cognition-mental-health.github.io/

Research recruiting participants: https://cognition-mental-health.github.io/projects
Posts Replies Media Videos
alexandrapike.bsky.social
Some cool preliminary research on ingroup bias and how multiple dimensions of identity influence these biases. Thanks @miweissflog.bsky.social!
October 27, 2025 at 4:00 PM
alexandrapike.bsky.social
Best title of the day goes to @jamiecockcroft.bsky.social - lovely presentation and figures and an honest account of accidental replication.
October 27, 2025 at 3:45 PM
alexandrapike.bsky.social
Tabea-Maria Haase then presented some really thought-provoking work on some maths heuristics in children - particularly the operational momentum effect, look it up if you’ve not heard about it!
October 27, 2025 at 3:42 PM
alexandrapike.bsky.social
@chloebrunskill.bsky.social presents in our shorter talk session on our memory for narratives - interesting to think about this other level of memory, rather than just items and episodes!
October 27, 2025 at 3:30 PM
alexandrapike.bsky.social
Wrapping up session 1 is @lewis-ball.bsky.social presenting cool work on the intersection between sleep, memory and social interaction - featuring cool Bayesian modelling of item and source memory.
October 27, 2025 at 2:30 PM
alexandrapike.bsky.social
Next up are funky brain pics: @dmwatson.bsky.social discussing how natural objects are represented in the brain.
October 27, 2025 at 1:48 PM
alexandrapike.bsky.social
First talk is by @emmacsullivan.bsky.social on the oft-discussed topic of sleep, smartphones and mental health. Nice to hear balanced evidence rather than just polarised opinions! (Example of polarised views in photo).
October 27, 2025 at 1:12 PM
alexandrapike.bsky.social
Super pleased that we are doing a @yorkpsychology.bsky.social Postdoc and RA showcase today! Thank you so much to @chloebrunskill.bsky.social and @arimoccia.bsky.social for doing all the hard work and bringing the humorous comics.
October 27, 2025 at 1:07 PM
alexandrapike.bsky.social
Enjoying hearing more about @kellydonegan.bsky.social’s argument that psychological studies need to move beyond the lab (and into apps) - courtesy of the Centre for Decision Science Summer School @royalholloway.bsky.social and @nurasidarus.bsky.social
Kelly presenting a slide with reasons for taking science beyond the lab.
September 11, 2025 at 10:50 AM
alexandrapike.bsky.social
Fun quick poster spotlight talk at #CPC2025 - come see me later at poster 2.19 if you want more info on what we did or to discuss our results! Preprint here: osf.io/preprints/ps... (picture cred to @kellydonegan.bsky.social who took this despite her own talk being about 5 mins later!)
Photo of me on a stage presenting
July 15, 2025 at 8:30 AM
alexandrapike.bsky.social
Assumption 4: parameters only weakly related to symptoms. Most of our pre-registered comparisons did not show relationships, and of those that did, only one of these was robust when estimating this effect within the model. In exploratory analyses there were a few small relationships, mean r=0.1.
a priori predicted relationships between parameters and clinical measures. A-C) correlation coefficients and 95% confidence intervals of relationships between parameters and either EFA latent variables, or questionnaire totals. D-E) relationships and confidence intervals for significant relationships from A-C. F-G) relationships when these were estimated by embedding the effect within the generative model - this is a central (quantile-based) posterior interval estimate from the MCMC draws, showing the 95% mass interval shaded, and the median as a solid line in the centre.
July 10, 2025 at 11:07 AM
alexandrapike.bsky.social
Assumption 3: however, we did not find robust generalisation between conceptually related parameters across task contexts. More details in paper and figure, but essentially, only inverse-temperature-like parameters were related - not learning rates.
Relationships between parameters. A) The correlation matrix of the relationship between parameters that were anticipated to reflect a ‘learning rate’ construct. B) The correlation matrix of the relationship between parameters that were anticipated to reflect an ‘inverse temperature’ construct. Colours represent strength of correlation. C) We estimated an ‘upper bound’ for the parameter-parameter relationship by embedding a correlation matrix within the model. The figure shows a central (quantile-based) posterior interval estimate from the MCMC draws, showing the 95% mass interval shaded, with the median as a solid line in the centre. D) Exploratory analyses of all possible parameter-parameter relationships (Bonferroni corrected for multiple comparisons, only significant correlations shown). Note that the significant relationships did not change when correcting for age and gender, except that there was a relationship between the reward bias reward sensitivity parameter, and the gamble inverse temperature parameter. E) The result of an exploratory factor analysis of the parameters across all tasks: a four-factor structure was found, with some ‘parameter constructs’ represented as factors. However, this structure had less evidence than a factor structure in which parameters were grouped by task, in a confirmatory factor analysis on task data from time 2. Tasks are described using a letter: ‘B’ for bandit, ‘Ef’ for effort, ‘Ex’ for explore, ‘G’ for gamble.
July 10, 2025 at 11:07 AM
alexandrapike.bsky.social
Assumption 2: test-retest reliability was good-to-excellent when estimated within the model itself (mean ICC=0.635), with the exception of the reward bias task.
Test-retest reliability. Each plot shows the mean and 95% confidence interval for the test-retest reliability of each parameter from the best-fitting model. ICC(C,1) refers to consistency intraclass correlation coefficient, ICC(A,1) to absolute ICC, and ‘joint model’ refers to the correlation coefficient estimated by embedding the correlation matrix within the model. Shaded areas between dashed lines indicate correlations suggestive of poor reliability (<0.4), fair reliability (0.4-0.6), good reliability (0.6-0.75) and excellent reliability (0.75-1.0).
July 10, 2025 at 11:07 AM
alexandrapike.bsky.social
Assumption 1: parameter recovery was generally excellent, other than reward sensitivity in the reward bias task.
Recovery of task parameters. Each point is the point estimate of a Pearson’s correlation between originally estimated parameters, and parameters recovered after synthetic data was generated using these fitted parameter values. Error bars represent the 95% confidence interval on the correlation estimate. All correlations were statistically significant. Note that the explore task model does not require fitting, hence there is an exact analytic solution and the recovery is perfect (correlation of r=1.0). Shaded areas between dashed lines indicate correlations suggestive of poor recovery (<0.4), fair recovery (0.4-0.6), good recovery (0.6-0.75) and excellent recovery (0.75-1.0). Parameter names are abbreviated: ‘lr’ is learning rate (‘lr+’ is a reward learning rate, and ‘lr-’ is punishment learning rate); ‘decay+’ is decay of positive values, and ‘decay-’ is decay of negative values; ‘sens’ is sensitivity (‘rewsens’ is reward sensitivity, ‘effsens’ is effort sensitivity, and ‘instsens’ is instruction sensitivity); ‘invtemp’ refers to inverse temperature; ‘acc’ is acceptance bias in the effort task; ‘loss’ is loss aversion, ‘risk+’ is risk aversion on gain trials, and ‘risk-’ is risk aversion on loss trials of the gamble task; ‘thetaV’ is value difference, and ‘thetaRU’ is relative uncertainty in the explore task; finally, ‘initV’ represents ‘initial value’ in the reward bias task.
July 10, 2025 at 11:07 AM
alexandrapike.bsky.social
We investigated a subset of these assumptions in the domain of reward processing in depression/anxiety/anhedonia. We recruited 548 participants online, and they completed five cognitive tasks and seven questionnaires. 115 of them were re-recruited after 14 days to re-complete the cognitive tasks.
Diagrams of the four tasks participants completed. A) Fluctuating bandit task. Each box shows an example screen: screens one and three show the start of two different trials, where all four doors are presented. Screens two and four show two example outcomes – a loss (screen two) and a gain and loss together (screen 4). B) Effort task. Participants classified numbers as even or odd, in sequences which varied in difficulty and provided varying reward. On the first screen, participants choose whether to accept or reject the offer to complete a round, with explicit information as to the difficulty of that round and the reward available. They are then presented with a number (screen two), classify it correctly so it turns green (screen three), and then are presented with another number (screen four). C) Explore task. Participants choose between two slot machines, labelled as risky or safe (screens one and three). Note that these could be in combination: risky-risky, safe-safe, or risky-safe. After making a choice, participants were presented with the number of points they obtained (screens two and four). D) Gamble task. Participants chose between one risky (shown on the right in this figure) and one safe (on the left) option on each trial. If they chose the safe option, they received that amount deterministically; otherwise, they received one of the two amounts listed on the ‘risky’ card with a 50/50 probability. Amounts could be a gain (green), loss (red), or neutral (0 points, grey). E) Reward bias task. Participants saw a random dot kinematogram, and selected the side corresponding to the motion of the dots. The slot machine on that side, if they chose correctly, then probabilistically delivered a reward (green ball) or no reward (grey ball).
July 10, 2025 at 11:07 AM
alexandrapike.bsky.social
After a failed small grant application), we then launched this project with some incredible Masters students (Rowan Board, Meggy Croal, Sabrina Lam), and the help of other v. talented postdocs in the lab (@eointravers.bsky.social and Vincent Valton) [All featured in this slightly bizarre jigsaw]
July 10, 2025 at 11:07 AM
alexandrapike.bsky.social
Really enjoying Prof Emma Robinson’s accessibly-delivered talk on behavioural markers of mood/antidepressant effects! Really clear justification for focus on behavioural readouts of mood - and v thought provoking notes on welfare. #BAP2025
Emma Robinson’s summary slide
June 30, 2025 at 9:02 AM
alexandrapike.bsky.social
Sneaking out for an early dinner in Manchester
June 29, 2025 at 4:16 PM
alexandrapike.bsky.social
Finally, the amazing Professor Camilla Nord presents on interoception - including some cool lived experience work on body sensations, and work on disgust and gastric rhythms, alongside work on energy conservation. #BAP2025
Camilla presenting on all of the exciting implications of her interoception work!
June 29, 2025 at 2:20 PM
alexandrapike.bsky.social
Dr Max Taquet, winner of the junior clinical award, presents on the use of electronic health records for drug repurposing, biomarker development and relapse prediction - including some really cool findings on symptom instability #BAP2025
Dr Taquet showing data suggesting that high instability in symptoms is a big predictor of hospitalisation.
June 29, 2025 at 2:06 PM
alexandrapike.bsky.social
My fab colleague Professor(!) Cairney @sacairney.bsky.social won the the Senior Translational Award #BAP2025 for his work on sleep deprivation and mental health - good thing these talks are now and not post the conference dinner.
Scott presenting a slide linking some of the many mental health conditions linked to sleep.
June 29, 2025 at 1:53 PM
alexandrapike.bsky.social
Next up is Dr Paris Lalousis, presenting really cool AI work that netted him the junior translational award. To me the emerging clusters maybe speak to immune vs non-immune symptoms? #BAP2025
Paris showing slides summarising a type of machine learning model
June 29, 2025 at 1:27 PM
alexandrapike.bsky.social
It’s #BAP2025 time! Kicking us off is Dr Megan Jackson, the winner of the junior non-clinical award talking about animal models of apathy and motivation. Key quote: ‘as scientists, we love a tight error bar’. @bapsych.bsky.social
Image of Megan Jackson presenting slides validating a novel apathy task
June 29, 2025 at 1:12 PM
alexandrapike.bsky.social
Super proud of my group today at the @yorkpsychology.bsky.social @yorkstjohn.bsky.social PGR conference! Amazing to actually have an eating disorder area of a poster session for once. @lucysta02475610.bsky.social @holmyers.bsky.social @lisawoodley.bsky.social
June 26, 2025 at 7:14 PM
alexandrapike.bsky.social
Upgraded my afternoon coffee. Not a bad day to work @yorkpsychology.bsky.social
June 2, 2025 at 3:33 PM