Biology subjects

Murawski, C.

Publications and source records attributed to Murawski, C..

9 recordsLinked to original sources

bioRxiv2025

When word order matters: human brains represent sentence meaning differently from large language models

Large language models based on the transformer architecture are now capable of producing human-like language. But do they encode and process linguistic meaning in a human-like way? Here, we address this question by analysing 7T fMRI data from 30 participants reading 108 sentences each. These sentences are carefully designed to disentangle sentence structure from word meaning, thereby testing whether transformer representations of sentence meaning resemble those formed by the brain. We found that while transformer models match brain representations better than models that completely ignore word order, all transformer models performed poorly overall. Further, transformers were significantly inferior to models explicitly designed to encode the structural relations between words. Our results provide insight into the nature of sentence representation in the brain, highlighting the critical role of sentence structure. They also cast doubt on the claim that transformers represent sentence meaning similarly to the human brain.

neuroscience↗

bioRxiv2025

Whom Do We Prefer to Learn From in Observational Reinforcement Learning?

Learning by observing others experiences is a hallmark of human intelligence. While the neurocomputational mechanisms underlying observational learning are well understood, less is known about whom people prefer to learn from in the context of observational learning. One hypothesis posits that learners prefer individuals who exhibit a high degree of decision noise (randomness in action selection), free riding on the costly exploration of others. An alternative hypothesis suggests that learners prefer individuals with low decision noise, and imitate consistent and reliable behavior. In a preregistered experiment, we found that most participants preferred to learn from low-noise individuals. Furthermore, exploratory analyses revealed that participants who preferred lownoise individuals tended to rely on imitation of others actions, whereas those who preferred high-noise individuals relied more on learning from others reward outcomes. These findings offer a potential computational account of how learning styles shape partner selection in social learning. Author SummaryIn our daily lives, we often learn by watching others. For example, when starting a new job, we might watch an experienced colleague to learn effective strategies, or an inexperienced coworker to avoid common mistakes. While previous studies have examined how people learn by observing others, less is known about how we decide whom to observe. This study explored whether people prefer to learn from individuals who make consistent choices or those who behave more randomly. At first glance, the answer seems obvious: we would naturally prefer to learn from those who make consistent, reliable decisions. However, there can also be value in learning from someone who behaves unpredictably. For example, when searching for a good restaurant, observing an adventurous friend who tries unfamiliar places might help us discover hidden gems. Despite this potential advantage, we found that most participants preferred consistent decision-makers. Further analysis revealed that participants who favored reliable partners tended to imitate their actions, while those who chose more exploratory partners focused more on outcomes. Our findings suggest that personal learning styles shape partner preferences. These insights could help us understand how people choose whom to learn from in everyday settings like classrooms, or workplaces.

animal behavior and cognition↗

bioRxiv2022

The neural dynamics associated with computational complexity

Many everyday tasks require people to solve computationally complex problems. However, little is known about the effects of computational hardness on the neural processes associated with solving such problems. Here, we draw on computational complexity theory to address this issue. We performed an experiment in which participants solved several instances of the 0-1 knapsack problem, a combinatorial optimization problem, while undergoing ultra-high field (7T) functional magnetic resonance imaging (fMRI). Instances varied in two task-independent measures of intrinsic computational hardness: complexity and proof hardness. We characterise a network of brain regions whose activation was correlated with both measures but in distinct ways, including the anterior insula, dorsal anterior cingulate cortex and the intra-parietal sulcus/angular gyrus. Activation and connectivity changed dynamically as a function of complexity and proof hardness, in line with theoretical computational requirements. Overall, our results suggest that computational complexity theory provides a suitable framework to study the effects of computational hardness on the neural processes associated with solving complex cognitive tasks.

neuroscience↗

bioRxiv2021

Task-independent metrics of computational hardness predict performance of human problem-solving

The survival of human organisms depends on our ability to solve complex tasks in the face of limited cognitive resources. However, little is known about the factors that drive the complexity of those tasks. Here, building on insights from computational complexity theory, we quantify the computational hardness of cognitive tasks using a set of task-independent metrics related to the computational resource requirements of individual instances of a task. We then examine the relation between those metrics and human behavior and find that they predict both time spent on a task as well as accuracy in three canonical cognitive tasks. Our findings demonstrate that performance in cognitive tasks can be predicted based on generic metrics of their inherent computational hardness. TeaserThe ability of humans to solve cognitive tasks is affected by generic mathematical properties of problems related to their computational complexity.

animal behavior and cognition↗

bioRxiv2020

Segment-Specific Optogenetic Stimulation in Drosophila melanogaster with Linear Arrays of Organic Light-Emitting Diodes

Optogenetics allows light-driven, non-contact control of neural systems, but light delivery remains challenging, in particular when fine spatial control of light is required to achieve local specificity. Here, we employ organic light-emitting diodes (OLEDs) that are micropatterned into linear arrays to obtain precise optogenetic control in Drosophila melanogaster larvae expressing the light-gated activator CsChrimson and the inhibitor GtACR2 within their peripheral sensory system. Our method allows confinement of light stimuli to within individual abdominal segments, which facilitates the study of larval behaviour in response to local sensory input. We show controlled triggering of specific crawling modes and find that targeted neurostimulation in abdominal segments switches the direction of crawling. More broadly, our work demonstrates how OLEDs can provide tailored patterns of light for photo-stimulation of neuronal networks, with future implications ranging from mapping neuronal connectivity in cultures to targeted photo-stimulation with pixelated OLED implants in vivo.

neuroscience↗

bioRxiv2018

Where the really hard choices are: A general framework to quantify decision difficulty

Life presents us with decisions of varying degrees of difficulty. Many of them are NP-hard, that is, they are computationally intractable. Two important questions arise: which properties of decisions drive extreme computational hardness and what are the effects of these properties on human-decision making? Here, we postulate that we can study the effects of computational complexity on human decision-making by studying the mathematical properties of individual instances of NP-hard problems. We draw on prior work in computational complexity theory, which suggests that computational difficulty can be characterized based on the features of instances of a problem. This study is the first to apply this approach to human decision-making. We measured hardness, first, based on typical-case complexity (TCC), a measure of average complexity of a random ensemble of instances, and, second, based on instance complexity (IC), a measure that captures the hardness of a single instance of a problem, regardless of the ensemble it came from. We tested the relation between these measures and (i) decision quality as well as (ii) time expended in a decision, using two variants of the 0-1 knapsack problem, a canonical and ubiquitous computational problem. We show that participants expended more time on instances with higher complexity but that decision quality was lower in those instances. These results suggest that computational complexity is an inherent property of the instances of a problem, which affect human and other kinds of computers.

animal behavior and cognition↗

bioRxiv2018

Hard decisions shape the neural coding of preferences

Hard decisions between equally valued alternatives can result in preference changes, meaning that subsequent valuations for chosen items increase and decrease for rejected items. Previous research suggests that this phenomenon is a consequence of cognitive dissonance reduction after the decision, induced by the mismatch between initial preferences and decision outcomes. In contrast, this functional magnetic resonance imaging and eye-tracking study tested whether preferences are already updated online while making decisions. Preference changes could be predicted from activity in left dorsolateral prefrontal cortex and precuneus during decision-making. Furthermore, fixation durations predicted both choice outcomes and subsequent preference changes. These preference adjustments became behaviourally relevant at re-evaluation, but only for choices that were remembered and were associated with hippocampus activity. Our findings refute classical explanations of post-choice dissonance reduction and instead suggest that preferences evolve dynamically as decisions arise, potentially as a mechanism to prevent stalemate situations in underdetermined decision scenarios.

neuroscience↗

bioRxiv2017

Electrophysiological indices reflect switches between Bayesian and heuristic strategies in perceptual learning

Belief updating entails the incorporation of new information about the environment into internal models of the world. Bayesian inference is the statistically optimal strategy for performing belief updating in the presence of uncertainty. An important open question is whether the use of cognitive strategies that implement Bayesian inference is dependent upon motivational state and, if so, how this is reflected in electrophysiological signatures of belief updating in the brain. Here we recorded the electroencephalogram of participants performing a simple reward learning task with both monetary and non-monetary instructive feedback conditions. Our aim was to distinguish the influence of the rewarding properties of feedback on belief updating from the information content of the feedback itself. A Bayesian updating model allowed us to quantify different aspects of belief updating across trials, including the size of belief updates and the uncertainty of beliefs. Faster learning rates were observed in the monetary feedback condition compared to the instructive feedback condition, while belief updates were generally larger, and belief uncertainty smaller, with monetary compared to instructive feedback. Larger amplitudes in the monetary feedback condition were found for three event-related potential components: the P3a, the feedback-related negativity (FRN) and the late positive potential (LPP). These findings suggest that motivational state influences inference strategies in reward learning, and this is reflected in the electrophysiological correlates of belief updating.

neuroscience↗

bioRxiv2017

The neural encoding of information prediction errors during non-instrumental information seeking

In a dynamic world, accurate beliefs about the environment are vital for survival, and individuals should therefore regularly seek out new information with which to update their beliefs. This aspect of behaviour is not well captured by standard theories of decision making, and the neural mechanisms of information seeking remain unclear. One recent theory posits that valuation of information results from representation of informative stimuli within canonical neural reward-processing circuits, even if that information lacks instrumental use. We investigated this question by recording EEG from twenty-three human participants performing a non-instrumental information-seeking task. In this task, participants could pay a monetary cost to receive advance information about the likelihood of receiving reward in a lottery at the end of each trial. Behavioural results showed that participants were willing to incur considerable monetary costs to acquire early but non-instrumental information. Analysis of the event-related potential elicited by informative cues revealed that the feedback-related negativity independently encoded both an information prediction error and a reward prediction error. These findings are consistent with the hypothesis that information seeking results from processing of information within neural reward circuits, and suggests that information may represent a distinct dimension of valuation in decision making under uncertainty.

neuroscience↗