About

Eric Schiffman, DDS, MS, is a Professor and the Director of Clinical Research in the School of Dentistry at the University of Minnesota. He has received over $19 million in research funding from the National Institutes of Health (NIH) and has authored over 60 peer-reviewed publications, 12 book chapters, and holds 3 patents with another pending. His past research as an NIH study principal investigator includes developing and publishing validated Diagnostic Criteria for Temporomandibular Disorders (DC/TMD) for the most common TMD, applicable in both clinical and research settings. He has also conducted research on TMD management, assessing the long-term effectiveness of medical management versus comprehensive rehabilitation with and without TMJ surgery in patients with symptomatic TMJ closed lock and limited mouth opening. Additionally, his work includes studying the longitudinal impact of intra-articular TMJ disorders on jaw pain, function, and disability. Schiffman has contributed to the NIH-funded National Dental Practice-Based Research Network by evaluating how dentists manage TMD patients in their practices. His recent projects include developing and clinically testing the Restful Jaw device, a device attached to dental chairs designed to support the jaw during various dental procedures, including third molar surgeries.

Research topics

• Computer Science
• Artificial Intelligence
• Political Science
• Machine Learning
• Computer Security
• Commerce
• Knowledge management
• Cognitive science
• Data science
• Mathematics
• Business
• Law
• Internet privacy
• Psychology
• Human–computer interaction
• Engineering

Selected publications

• A Surgeon's Guide to Machine Learning.

  UNC Libraries · 2025-06-21

  articleOpen accessSenior author

  Machine learning (ML) represents a collection of advanced data modeling techniques beyond the traditional statistical models and tests with which most clinicians are familiar. While a subset of artificial intelligence, ML is far from the science fiction impression frequently associated with AI. At its most basic, ML is about pattern finding, sometimes with complex algorithms. The advanced mathematical modeling of ML is seeing expanding use throughout healthcare and increasingly in the day-to-day practice of surgeons. As with any new technique or technology, a basic understanding of principles, applications, and limitations are essential for appropriate implementation. This primer is intended to provide the surgical reader an accelerated introduction to applied ML and considerations in potential research applications or the review of publications, including ML techniques.

  PublisherDOI

• Provable Distributional Value Iteration under Partial Observability

  ArXiv.org · 2025-05-10

  preprintOpen accessSenior author

  In many real-world planning tasks, agents must tackle uncertainty about the environment's state and variability in the outcomes induced by stochastic dynamics and rewards. Motivated by recent progress in world model approaches, where latent models approximate beliefs and support planning, we extend Distributional Reinforcement Learning (DistRL), which models the entire return distribution for fully observable domains, to Partially Observable Markov Decision Processes (POMDPs). Concretely, we introduce new distributional Bellman operators for partial observability and prove their convergence under the supremum p-Wasserstein metric. We also propose a finite representation of these return distributions via psi-vectors, generalizing the classical alpha-vectors in POMDP solvers. Building on this, we develop Distributional Point-Based Value Iteration (DPBVI), which integrates psi-vectors into a standard point-based backup procedure, bridging DistRL and POMDP planning. Our experiments demonstrate that DPBVI recovers classical Point-Based Value Iteration (PBVI) in the risk-neutral case, validating the distributional extension.

  PublisherOA PDFDOI

• Robust Parkinson’s Disease Screening from Speech via Self-Supervised Learning Models and Task-Wise Analysis

  2025-12-15

  article1st authorCorresponding

  Changes in speech often precede motor symptoms in Parkinson’s disease, offering a scalable and non-invasive biomarker for early diagnosis. In this study, frozen self-supervised speech encoders, namely Wav2Vec2.0, HuBERT, and WavLM, are leveraged to extract acoustic and phonetic features. These features are fed to a lightweight Bidirectional Gated Recurrent Unit, with attention pooling that focuses on clinically salient segments. Using speaker-independent 10-fold cross-validation on each dataset and a speaker-held-out test set for final evaluation, this study explores both overall and task-wise performance on two languages, Italian and Spanish. On Italian speech, HuBERT attains 97.9% accuracy with AUC = 99.0, approaching screening quality, while on PC-GITA, WavLM generalizes best with 88.4% accuracy and AUC = 95.8, underscoring the role of language/domain in backbone choice. Task-wise analysis shows Reading/Sentences are the most reliable, high-yield probes, with DDK also robust; Vowels/Monologue are more variable and serve as complementary evidence. The proposed pipeline is compute-efficient, and well-suited for scalable clinical pre-screening of Parkinson’s disease.

  PublisherDOI

• Islamic Chatbots in the Age of Large Language Models

  ArXiv.org · 2025-12-31

  articleOpen access1st authorCorresponding

  Large Language Models (LLMs) are rapidly transforming how communities access, interpret, and circulate knowledge, and religious communities are no exception. Chatbots powered by LLMs are beginning to reshape authority, pedagogy, and everyday religious practice in Muslim communities. We analyze the landscape of LLM powered Islamic chatbots and how they are transforming Islamic religious practices e.g., democratizing access to religious knowledge but also running the risk of erosion of authority. We discuss what kind of challenges do these systems raise for Muslim communities and explore recommendations for the responsible design of these systems.

  PublisherOA PDF

• Algorithmic Fairness in AI Surrogates for End-of-Life Decision-Making

  ArXiv.org · 2025-10-16

  preprintOpen access1st authorCorresponding

  Artificial intelligence surrogates are systems designed to infer preferences when individuals lose decision-making capacity. Fairness in such systems is a domain that has been insufficiently explored. Traditional algorithmic fairness frameworks are insufficient for contexts where decisions are relational, existential, and culturally diverse. This paper explores an ethical framework for algorithmic fairness in AI surrogates by mapping major fairness notions onto potential real-world end-of-life scenarios. It then examines fairness across moral traditions. The authors argue that fairness in this domain extends beyond parity of outcomes to encompass moral representation, fidelity to the patient's values, relationships, and worldview.

  PublisherOA PDFDOI

• Islamic Chatbots in the Age of Large Language Models

  arXiv (Cornell University) · 2025-12-31

  preprintOpen access1st authorCorresponding

  Large Language Models (LLMs) are rapidly transforming how communities access, interpret, and circulate knowledge, and religious communities are no exception. Chatbots powered by LLMs are beginning to reshape authority, pedagogy, and everyday religious practice in Muslim communities. We analyze the landscape of LLM powered Islamic chatbots and how they are transforming Islamic religious practices e.g., democratizing access to religious knowledge but also running the risk of erosion of authority. We discuss what kind of challenges do these systems raise for Muslim communities and explore recommendations for the responsible design of these systems.

  PublisherDOI

• QuantX: A Framework for Hardware-Aware Quantization of Generative AI Workloads

  ArXiv.org · 2025-05-12

  preprintOpen access1st authorCorresponding

  We present QuantX: a tailored suite of recipes for LLM and VLM quantization. It is capable of quantizing down to 3-bit resolutions with minimal loss in performance. The quantization strategies in QuantX take into account hardware-specific constraints to achieve efficient dequantization during inference ensuring flexible trade-off between runtime speed, memory requirement and model accuracy. Our results demonstrate that QuantX achieves performance within 6% of the unquantized model for LlaVa-v1.6 quantized down to 3-bits for multiple end user tasks and outperforms recently published state-of-the-art quantization techniques. We further integrate one particular technique from QuantX into the popular Llama.cpp framework and show its feasibility in terms of runtime compared to the mainstream quantization techniques from Llama.cpp. Lastly, this manuscript provides insights into the LLM quantization process that motivated the range of recipes and options that are incorporated in QuantX.

  PublisherOA PDFDOI

• Building Personality-Adaptive Conversational AI for Mental Health Therapy

  2024-11-22 · 5 citations

  article

  Many people with mental health problems cannot get professional help for various reasons such as lack of awareness, unavailability, unaffordability, etc. A virtual conversational agent can offer an alternative to deliver mental health care that is accessible, affordable, and scalable. However, building such agents using a one-size-fits-all approach may not be effective for everyone, as different individuals have different personality types that dictate how they communicate with chatbots. Therefore, developing therapy chatbots that can adjust to the user's personality is important. In this work, we present the important role of personality-adaptive conversational agents (PACAs) in the context of mental healthcare. We designed an architecture around traditional machine learning (ML) models and open-source large language models (LLMs) to build a PACA for mental health therapy, developed a working prototype based on it, and conducted a user study to conclude that personality-adaptiveness is indeed an important feature for mental health chatbots.

  PublisherDOI

• HIV-Associated Pseudoaneurysms: A Comprehensive Review

  Cureus · 2024-10-21

  reviewOpen access

  A pseudoaneurysm (PSA) is a contained vascular rupture that typically occurs following catheterization, at the anastomotic site between a native artery and a synthetic graft, post-trauma, or as a result of infection. It is characterized by a hematoma surrounded by tissue, often emerging as a complication of invasive arterial interventions. In patients with HIV/AIDS, PSAs can develop due to vessel wall disruption caused by chronic inflammation, opportunistic infections (such as cytomegalovirus or tuberculosis), or the direct effects of the virus, leading to abnormal blood flow into a chamber confined by adjacent tissue. The clinical presentation of PSAs varies based on their size and location. Diagnosis can be achieved through ultrasonography with color Doppler, contrast-enhanced computed tomography (CT), magnetic resonance angiography (MRA), or digital subtraction angiography (DSA). Treatment modalities include surgery, ultrasound-guided compression, thrombin injection, and endovascular techniques. This review discusses the pathophysiology, histology, diagnosis, and therapeutic options for HIV-related PSAs. Additionally, risk factors and rare complications associated with PSAs are explored in detail.

  PublisherOA PDFDOI

• E ² T ² : Emote Embedding for <i>T</i> witch <i>T</i> oxicity Detection

  2024-11-11

  articleOpen access

  The Internet has become the medium of choice for socialization and communication. The rise of live streaming services has created countless online communities of varying sizes with their own jokes, references, slang, and other means of communication. One of the largest live streaming services is Twitch.tv or Twitch, where a unique culture of niche language and emote usage has developed. Emotes are custom-made images, or GIFs, used in chat with varying degrees of access influenced by channel and external site subscription status. Emotes render standard forms of English Natural Language Process- ing (NLP) for tasks such as detection of toxicity or cyberbullying ineffective on Twitch. In this paper, we propose a methodology and offer a largely-trained dataset for detecting emote-based toxicity on live streaming platforms such as Twitch.

  PublisherOA PDFDOI

Frequent coauthors

• Jaideep Srivastava

  26 shared

• Carly Eckert

  24 shared

• Ankur Teredesai

  22 shared

• Noshir Contractor

  11 shared

• Brian Keegan

  University of Colorado Boulder

  10 shared

• Dmitri Williams

  University of Exeter

  9 shared

• Marshall Scott Poole

  7 shared

• Zoheb Borbora

  University of Minnesota System

  6 shared

Education

• PhD, Computer Science

  University of Minnesota

• Bachelors of Science, Computer Science

  Rochester Institute of Technology

  2006

Awards & honors

• The National Dental PBRN (2014 - 2019)
• Proximal caries and bone loss detection accuracy using i (20…
• NIH NIDCR NATL INST OF DENTAL (2011 - 2015)