Attentional Modulation of Neural Activity in Multi-Attribute Decision Making

/Abstract Decision making is a core cognitive process underlying myriad behaviors in day-to-day life. While great strides have been made in uncovering elements of the neural processes underlying decision making, key questions remain, especially related to the type of complex multi-attribute decisions that are common in the real world. Our novel task isolates each piece of information related to the value of options and requires subjects to view these attributes of options one at a time. Because of this design, we can monitor attention throughout and study the neural responses at each stage. This task also presents the distinct advantage of being viable for both human subjects and non-human primates. Already, we have gained key insights from recordings of neuronal activity in the pre-supplementary motor area (preSMA) of macaques performing the task. We found that preSMA neurons encode action value signals that accumulate information about the options. We also discovered that the focus of attention plays a key role in the value estimation process in these neurons. Attention both uniformly enhances the activity of neurons when the option in their preferred spatial position is attended and increases the gain of value representation in these neurons. With fundamentally the same task, we have collected data from patients undergoing intracranial recordings as part of their treatment for medically intractable focal epilepsy. With broad coverage of the brain across the population of patients, we have been able to identify a number of brain regions involved in different aspects of the task. Now, we can combine the advantages of both data sets to make additional progress. First, in Aim 1, I will implement a computational model for Attention Modulated Multi- Attribute Decisions (AM-MAD) consistent with the observed activity patterns in the preSMA data. This model will allow us to test the hypothesis that the observed forms of attentional modulation of value representation serve to prevent premature choices by allowing attended options to overcome inhibition. In Aim 2, broader networks for value estimation and option selection will be identified from the human sEEG data using a set of dimensionality reduction tools. This approach should reveal whether similar attentional modulation of value representation is present in corresponding areas in humans, as well as how these factors influence activity in other areas. Finally, in Aim 3, I will use new tools that have been developed for causal inference and are ideally suited to multi- channel intracranial recordings. This will allow us to study the flow of information through large-scale networks in the brain and test whether regions that show activity changes related to specific aspects of the task are directly interacting, e.g. whether activity in premotor areas is driven by both attention- and value-encoding frontal areas. With this additional information, the AM-MAD model will be refined with more detail about value and attention- related inputs. Together, this will serve to both advance our understanding of the neural mechanisms of attention in multi-attribute decision making and establish a set of tools that will be broadly applicable for studying cognitive processes across species and recording modalities. Project Number: 1K99MH139728-01 | Fiscal Year: 2026 | NIH Institute/Center: National Institute of Mental Health (NIMH) | Principal Investigator: Aaron Sampson | Institution: JOHNS HOPKINS UNIVERSITY, BALTIMORE, MD | Award Amount: $104,349 | Activity Code: K99 | Study Section: Special Emphasis Panel[ZMH1 ERB-G (01)] View on NIH RePORTER: https://reporter.nih.gov/project-details/11116500