Michael J. Kahana · University of Pennsylvania · PhdFit
Resume-aware faculty matching
Find professors who actually fit you
Upload your resume. PhdFit's six research agents compare your background with faculty profiles, recent publications, lab focus, and outreach opportunities, then rank professors with evidence you can review.
Resume-aware matchingEvidence from recent researchOutreach drafts included
Advisor Match Brief Resume uploaded
Resume signals5 found
NLPHCIFairnessGraph MLPyTorch
Six-agent analysisRunning
Axel · Candidate AnalystResume parsed
Nova · Professor ResearcherRecent papers
Echo · Match ExplainerEvidence linked
Atlas · Strategy AdvisorFit ranked
Lyra · Outreach WriterDraft queued
Juno · Meeting Prep CompilerNext step ready
SC
Dr. Sarah Chen
Stanford · NLP & interpretability
91
fit
Fairness + NLP projects overlap with her recent work.
Strong methods fit: model evaluation and interpretability.
Clear outreach angle for current trustworthy AI research.
Outreach angle
Ask how her lab is extending interpretability methods into fairness audits for real-world AI systems.
Michael J. Kahana, Ph.D, is a faculty member in the Biomedical Graduate Studies at the Perelman School of Medicine at the University of Pennsylvania. His research focuses on human episodic memory for verbal, visual, and spatial information. He conducts experiments that measure behavioral and electrophysiological responses during memory tasks and develops computational models to explain the resulting data. His lab is one of several studying the electrophysiological responses of neurons through direct intracranial electroencephalographic (iEEG) recording from the living human brain, often involving epilepsy patients with surgically implanted electrodes. By analyzing brain activity, including responses of individual neurons, in relation to task variables, he studies the neurophysiological basis of memory with high spatial and temporal resolution. Current projects include studies of spatial navigation using a virtual taxi driver game and computational modeling of the role of temporal context in visual and verbal memory.
bioRxiv (Cold Spring Harbor Laboratory) · 2026-03-19
articleOpen access
Abstract Intracranial stimulation can enhance episodic memory in humans; however, the behavioral effects vary substantially across individuals and stimulation sites. Here, we investigated whether the network embedding of a stimulation target, defined by MRI-based normative structural and functional connectivity, accounts for variability in stimulation-linked memory enhancement. We analyzed data from 50 adults with medically refractory epilepsy who underwent intracranial EEG monitoring and completed a verbal delayed free-recall task during stimulation of left temporal cortex sites across 61 sessions (39 closed-loop; 22 random). On average, closed-loop stimulation delivered during classifier-detected low-encoding states increased recall rates, whereas random stimulation produced no reliable benefit. Diffusion tractography from a normative database showed that sites yielding greater memory enhancement were characterized by stronger structural coupling to a distributed fronto-temporo-parietal network. Greater structure-function congruence with a normative verbal-encoding activation network predicted larger closed-loop memory benefit (Spearman ρ = 0.58, P < 0.0001). Functional connectivity exhibited overlapping trends but did not yield robust regional associations after permutation correction. Multivariate Partial Least Squares Structural Equation Modeling further identified stimulation mode, baseline memory, and a structural profile factor as independent predictors of memory enhancement, with no independent contribution of functional connectivity. These findings indicate that reliable stimulation-driven memory improvement depends not only on the timing of stimulation, but also on whether the stimulated target is structurally embedded within an encoding-relevant network scaffold. Significance statement Memory enhancement through direct brain stimulation holds substantial clinical promise, yet inconsistent outcomes have limited its therapeutic translation. This study shows that the effectiveness of closed-loop brain stimulation for memory improvement is determined by the structural network architecture of the stimulation target. Sites more deeply embedded within white-matter pathways connecting a distributed verbal encoding network yield the greatest mnemonic benefits when stimulation is delivered adaptively during poor encoding states. These findings establish a principled, network-based rationale for precision-guided neuromodulation: optimizing both the target’s structural embedding and the timing of stimulation delivery are necessary and complementary conditions for reliable, individualized memory enhancement.
Abstract Despite promising results, it remains unclear how to optimally target and personalize closed-loop stimulation to ameliorate deficits in memory and other cognitive functions. We hypothesized that evoked connectivity – the measurement of neural pathway activation using single pulses of electrical stimulation – can guide patient-specific selection of stimulation location and parameters for memory. We characterized brain-wide evoked connectivity profiles of stimulation in memory-related brain regions recorded from 81 patients undergoing intracranial monitoring for epilepsy, showing that greater evoked connectivity between the lateral temporal cortex and the broader memory network (including the mesial temporal lobe, limbic regions and prefrontal cortex) corroborates observations of memory improvement by lateral temporal cortex stimulation. We first found that the lateral temporal cortex, compared to other stimulated regions, evokes the greatest and most distributed connectivity response throughout other memory-related regions. Evoked connectivity in downstream regions is greatest when stimulating a previously identified optimal target for memory improvement, bordering white matter at the rostrocaudal center of the middle temporal gyrus. Evoked connectivity corroborates other biomarkers of memory improvement by closed-loop stimulation, including increased resting-state functional connectivity between stimulated and recorded sites and increased modulation of oscillatory power. These results provide insight into the network mechanisms of stimulation for memory and suggest that evoked connectivity can more broadly predict the functional effects of closed-loop stimulation to prospectively guide targeting and parameter selection.
A reconstruction-of-order task illuminated the dynamics and strategies that underlie serial order recall. An initial benchmark experiment, either with no variation in spatial positions or with spatial positions coinciding with temporal positions, yielded bowed symmetrical serial position functions in each case, consistent with both simple chaining and simple positional coding models. In contrast, these simple models were challenged by two additional experiments, which orthogonally varied temporal and spatial positions. These experiments yielded large performance differences between recalling temporal and spatial information. In the temporal condition, participants attempted to reconstruct the temporal order of words that were positioned alphabetically within a vertical array. In the spatial condition, participants attempted to reconstruct the spatial positions of words presented in a temporal sequence based on their alphabetical order. After viewing each list, all the words appeared alphabetically, and participants reconstructed the order of the words according to their instructed condition. Compared to temporal recall, spatial recall exhibited superior performance and a more bowed symmetrical serial position function. Analyses showed the effects of temporal contiguity in the spatial condition and spatial contiguity in the temporal condition. These findings suggest the theoretical conclusion that participants do not focus on the words' identities but rather on the temporal-spatial pattern in which the words occur during the study display (i.e., the temporal sequence of the spatial locations in which the words are shown).
Human and animal studies implicate theta and alpha oscillations in memory function. We tested whether theta, alpha, and beta phase consistency predicts memory encoding dynamics in neurosurgical patients performing delayed free recall tasks with either unrelated ( N = 188: 99 male, 89 female) or categorized words ( N = 157: 88 male, 69 female). We observed widespread post-stimulus phase consistency (3–21 Hz) and, crucially, identified distinct frequency-specific patterns predictive of successful encoding. Specifically, increased early list item recall was significantly correlated across subjects with increased theta (3–7 Hz) phase consistency. Subsequent recall analyses, controlling for serial position, revealed distinct frequency signatures for successfully encoded items: theta (3–6 Hz) and alpha (9–14 Hz) for unrelated lists, and theta (3–6 Hz) and beta (14–19 Hz) for categorized lists. Regional analyses for unrelated lists highlighted the lateral temporal cortex for theta effects and the prefrontal cortex for both theta and alpha consistency. These findings provide novel evidence for the frequency-specific presence of increased phase consistency during episodic encoding, revealing its sensitivity to both item context and temporal position within a learning sequence.
The present study investigates whether electroencephalogram activity reflects age-related memory changes during encoding. We recorded scalp electroencephalogram in 151 young adults (aged 18-30) and 37 older adults (aged 60-85) as they memorized lists of words. Participants studied the word lists either under full attention or while performing a secondary task that required them to make semantic judgments about each word. Although the secondary task reduced recall among all participants, differences in recall performance between the age groups were smaller when participants performed a secondary task at encoding. Older adults also exhibited distinct neural subsequent memory effects, characterized by less negativity in the alpha frequencies compared to young adults. Multivariate classifiers trained on neural features successfully predicted subsequent memory at the trial level in both young and older adults, and captured the differential effects of task demands on memory performance between young and older adults. The findings indicate that neural biomarkers of successful memory vary with both cognitive aging and task demands. (PsycInfo Database Record (c) 2025 APA, all rights reserved).
OBJECTIVE: Closed-loop neuromodulatory therapies require devices that can decode ongoing brain states and deliver multi-site stimulation. METHODS: We describe the Smart Neurostimulation System (SNS), a cranially mounted implant with 60 configurable recording/stimulation channels, inductive power, and onboard spectral-feature classification. In three freely-moving sheep, we streamed local-field potentials and conducted two parameter-sweep experiments. RESULTS: Cross-validated movement classifiers achieved an average AUC exceeding 0.95. Increasing stimulation amplitude and frequency produced post-stimulation elevations in α-band (8-12 Hz) and γ-band (78-82 Hz) power at most target locations. CONCLUSION: The SNS unifies high-density sensing, real-time brain state decoding, and programmable closed-loop stimulation in a single device, demonstrating behavioral-state prediction and parameter-dependent neuromodulation in vivo. SIGNIFICANCE: These findings establish a preclinical foundation for biomarker-guided stimulation targeting distributed cortical networks underlying memory and cognition.
When individuals repeatedly study and recall information across multiple learning trials their responses exhibit increasing levels of subjective organization. Whereas classic studies investigated the evolution of organization across lists within the short-time span of a single session, here we ask how memory changes over many days. Specifically, we examine how semantic, temporal, and subjective organization during a recall period shapes memory after days of intervening cognitive activity. Analyzing data from two multi-session free recall experiments, we find that subjects demonstrate a strong tendency to cluster recalls based on previous output order, with this effect strengthening across sessions. In line with the idea that thoughts become memories, we show that even false memories produced on a given session tend to re-occur on subsequent days. Our results attest to the crucial role that retrieval plays in shaping long-term episodic memory.
We propose a comprehensive model of how experiences are encoded and retrieved from memory. At the core of the model is a dynamic retrieval process incorporating two essential mechanisms: iterative retrieval, whereby information is sequentially sampled from memory to access the full history of experiences; and competitive retrieval, whereby the most prominent features in memory inhibit the recollection of other features. Together with context-based encoding, the model quantitatively explains well-known facts about response order and inter-response times in recall experiments. We show that our retrieval process maps closely to existing decision frameworks, such as drift–diffusion models, suggesting that the memory system plays a fundamental role in a wide-ranging set of decision-making settings. • We propose a model of encoding, retrieval, and response incorporating rich memory dynamics. • Information is sequentially sampled from memory and competes for retrieval. • The model can explain recall and response-time data in free-recall experiments. • The model sheds light on the role of memory in established decision frameworks, including drift-diffusion models.
