Artificial intelligence holds great promise for expanding access to expert medical knowledge and reasoning. However, most evaluations of language models rely on static vignettes and multiple-choice questions that fail to reflect the complexity and nuance of evidence-based medicine in real-world settings. In clinical practice, physicians iteratively formulate and revise diagnostic hypotheses, adapting each subsequent question and test to what they've just learned, and weigh the evolving evidence before committing to a final diagnosis. To emulate this iterative process, we introduce the Sequential Diagnosis Benchmark, which transforms 304 diagnostically challenging New England Journal of Medicine clinicopathological conference (NEJM-CPC) cases into stepwise diagnostic encounters. A physician or AI begins with a short case abstract and must iteratively request additional details from a gatekeeper model that reveals findings only when explicitly queried. Performance is assessed not just by diagnostic accuracy but also by the cost of physician visits and tests performed. We also present the MAI Diagnostic Orchestrator (MAI-DxO), a model-agnostic orchestrator that simulates a panel of physicians, proposes likely differential diagnoses and strategically selects high-value, cost-effective tests. When paired with OpenAI's o3 model, MAI-DxO achieves 80% diagnostic accuracy--four times higher than the 20% average of generalist physicians. MAI-DxO also reduces diagnostic costs by 20% compared to physicians, and 70% compared to off-the-shelf o3. When configured for maximum accuracy, MAI-DxO achieves 85.5% accuracy. These performance gains with MAI-DxO generalize across models from the OpenAI, Gemini, Claude, Grok, DeepSeek, and Llama families. We highlight how AI systems, when guided to think iteratively and act judiciously, can advance diagnostic precision and cost-effectiveness in clinical care.

In the modern world, we are permanently using, leveraging, interacting with, and relying upon systems of ever higher sophistication, ranging from our cars, recommender systems in e-commerce, and networks when we go online, to integrated circuits when using our PCs and smartphones, the power grid to ensure our energy supply, security-critical software when accessing our bank accounts, and spreadsheets for financial planning and decision making. The complexity of these systems coupled with our high dependency on them implies both a non-negligible likelihood of system failures, and a high potential that such failures have significant negative effects on our everyday life. For that reason, it is a vital requirement to keep the harm of emerging failures to a minimum, which means minimizing the system downtime as well as the cost of system repair. This is where model-based diagnosis comes into play. Model-based diagnosis is a principled, domain-independent approach that can be generally applied to troubleshoot systems of a wide variety of types, including all the ones mentioned above, and many more. It exploits and orchestrates i.a. techniques for knowledge representation, automated reasoning, heuristic problem solving, intelligent search, optimization, stochastics, statistics, decision making under uncertainty, machine learning, as well as calculus, combinatorics and set theory to detect, localize, and fix faults in abnormally behaving systems. In this thesis, we will give an introduction to the topic of model-based diagnosis, point out the major challenges in the field, and discuss a selection of approaches from our research addressing these issues.

This work proposes a taxonomy for diagnosis computation methods which allows their standardized assessment, classification and comparison. The aim is to (i) give researchers and practitioners an impression of the diverse landscape of available diagnostic techniques, (ii) allow them to easily retrieve the main features as well as pros and cons of the approaches, (iii) enable an easy and clear comparison of the techniques based on their characteristics wrt. a list of important and well-defined properties, and (iv) facilitate the selection of the "right" algorithm to adopt for a particular problem case, e.g., in practical diagnostic settings, for comparison in experimental evaluations, or for reuse, modification, extension, or improvement in the course of research.

Reiter's HS-Tree is one of the most popular diagnostic search algorithms due to its desirable properties and general applicability. In sequential diagnosis, where the addressed diagnosis problem is subject to successive change through the acquisition of additional knowledge about the diagnosed system, HS-Tree is used in a stateless fashion. That is, the existing search tree is discarded when new knowledge is obtained, albeit often large parts of the tree are still relevant and have to be rebuilt in the next iteration, involving redundant operations and costly reasoner calls. As a remedy, we propose DynamicHS, a variant of HS-Tree that avoids these redundancy issues by maintaining state throughout sequential diagnosis while preserving all desirable properties of HS-Tree. Evaluations in a problem domain where HS-Tree is the state-of-the-art diagnostic method reveal stable and significant time savings achieved by DynamicHS.

Sequential diagnosis prediction on the Electronic Health Record (EHR) has been proven crucial for predictive analytics in the medical domain. EHR data, sequential records of a patient's interactions with healthcare systems, has numerous inherent characteristics of temporality, irregularity and data insufficiency. Some recent works train healthcare predictive models by making use of sequential information in EHR data, but they are vulnerable to irregular, temporal EHR data with the states of admission/discharge from hospital, and insufficient data. To mitigate this, we propose an end-to-end robust transformer-based model called SETOR, which exploits neural ordinary differential equation to handle both irregular intervals between a patient's visits with admitted timestamps and length of stay in each visit, to alleviate the limitation of insufficient data by integrating medical ontology, and to capture the dependencies between the patient's visits by employing multi-layer transformer blocks. Experiments conducted on two real-world healthcare datasets show that, our sequential diagnoses prediction model SETOR not only achieves better predictive results than previous state-of-the-art approaches, irrespective of sufficient or insufficient training data, but also derives more interpretable embeddings of medical codes. The experimental codes are available at the GitHub repository (https://github.com/Xueping/SETOR).

Reiter's HS-Tree is one of the most popular diagnostic search algorithms due to its desirable properties and general applicability. In sequential diagnosis, where the addressed diagnosis problem is subject to successive change through the acquisition of additional knowledge about the diagnosed system, HS-Tree is used in a stateless fashion. That is, the existing search tree is discarded when new knowledge is obtained, albeit often large parts of the tree are still relevant and have to be rebuilt in the next iteration, involving redundant operations and costly reasoner calls. As a remedy to this, we propose DynamicHS, a variant of HS-Tree that avoids these redundancy issues by maintaining state throughout sequential diagnosis while preserving all desirable properties of HS-Tree. Preliminary results of ongoing evaluations in a problem domain where HS-Tree is the state-of-the-art diagnostic method suggest significant time savings achieved by DynamicHS by reducing expensive reasoner calls.

Given a malfunctioning system, sequential diagnosis aims at identifying the root cause of the failure in terms of abnormally behaving system components. As initial system observations usually do not suffice to deterministically pin down just one explanation of the system's misbehavior, additional system measurements can help to differentiate between possible explanations. The goal is to restrict the space of explanations until there is only one (highly probable) explanation left. To achieve this with a minimal-cost set of measurements, various (active learning) heuristics for selecting the best next measurement have been proposed. We report preliminary results of extensive ongoing experiments with a set of selection heuristics on real-world diagnosis cases. In particular, we try to answer questions such as "Is some heuristic always superior to all others?", "On which factors does the (relative) performance of the particular heuristics depend?" or "Under which circumstances should I use which heuristic?"

When a system behaves abnormally, sequential diagnosis takes a sequence of measurements of the system until the faults causing the abnormality are identified, and the goal is to reduce the diagnostic cost, defined here as the number of measurements. To propose measurement points, previous work employs a heuristic based on reducing the entropy over a computed set of diagnoses. This approach generally has good performance in terms of diagnostic cost, but can fail to diagnose large systems when the set of diagnoses is too large. Focusing on a smaller set of probable diagnoses scales the approach but generally leads to increased average diagnostic costs. In this paper, we propose a new diagnostic framework employing four new techniques, which scales to much larger systems with good performance in terms of diagnostic cost. First, we propose a new heuristic for measurement point selection that can be computed efficiently, without requiring the set of diagnoses, once the system is modeled as a Bayesian network and compiled into a logical form known as d-DNNF. Second, we extend hierarchical diagnosis, a technique based on system abstraction from our previous work, to handle probabilities so that it can be applied to sequential diagnosis to allow larger systems to be diagnosed. Third, for the largest systems where even hierarchical diagnosis fails, we propose a novel method that converts the system into one that has a smaller abstraction and whose diagnoses form a superset of those of the original system; the new system can then be diagnosed and the result mapped back to the original system. Finally, we propose a novel cost estimation function which can be used to choose an abstraction of the system that is more likely to provide optimal average cost. Experiments with ISCAS-85 benchmark circuits indicate that our approach scales to all circuits in the suite except one that has a flat structure not susceptible to useful abstraction.

Model-based diagnostic reasoning often leads to a large number of diagnostic hypotheses. The set of diagnoses can be reduced by taking into account extra observations (passive monitoring), measuring additional variables (probing) or executing additional tests (sequential diagnosis/test sequencing). In this paper we combine the above approaches with techniques from Automated Test Pattern Generation (ATPG) and Model-Based Diagnosis (MBD) into a framework called FRACTAL (FRamework for ACtive Testing ALgorithms). Apart from the inputs and outputs that connect a system to its environment, in active testing we consider additional input variables to which a sequence of test vectors can be supplied. We address the computationally hard problem of computing optimal control assignments (as defined in FRACTAL) in terms of a greedy approximation algorithm called FRACTAL-G. We compare the decrease in the number of remaining minimal cardinality diagnoses of FRACTAL-G to that of two more FRACTAL algorithms: FRACTAL-ATPG and FRACTAL-P. FRACTAL-ATPG is based on ATPG and sequential diagnosis while FRACTAL-P is based on probing and, although not an active testing algorithm, provides a baseline for comparing the lower bound on the number of reachable diagnoses for the FRACTAL algorithms. We empirically evaluate the trade-offs of the three FRACTAL algorithms by performing extensive experimentation on the ISCAS85/74XXX benchmark of combinational circuits.

Sequential diagnosis takes measurements of an abnormal system to identify faulty components, where the goal is to reduce the diagnostic cost , defined here as the number of measurements. To propose measurement points, previous work employs a heuristic based on reducing the entropy over a set of diagnoses , which can be impractical when the set of diagnoses is too large. Focusing on a smaller set of probable diagnoses scales the approach but generally leads to increased diagnostic cost. We propose a new diagnostic framework employing three new techniques — a more efficient heuristic for measurement point selection, abstraction-based sequential diagnosis, and component cloning — which scales to large systems with good performance in terms of diagnostic cost.