-
Human-AI collectives produce the most accurate differential diagnoses
Authors:
N. Zöller,
J. Berger,
I. Lin,
N. Fu,
J. Komarneni,
G. Barabucci,
K. Laskowski,
V. Shia,
B. Harack,
E. A. Chu,
V. Trianni,
R. H. J. M. Kurvers,
S. M. Herzog
Abstract:
Artificial intelligence systems, particularly large language models (LLMs), are increasingly being employed in high-stakes decisions that impact both individuals and society at large, often without adequate safeguards to ensure safety, quality, and equity. Yet LLMs hallucinate, lack common sense, and are biased - shortcomings that may reflect LLMs' inherent limitations and thus may not be remedied…
▽ More
Artificial intelligence systems, particularly large language models (LLMs), are increasingly being employed in high-stakes decisions that impact both individuals and society at large, often without adequate safeguards to ensure safety, quality, and equity. Yet LLMs hallucinate, lack common sense, and are biased - shortcomings that may reflect LLMs' inherent limitations and thus may not be remedied by more sophisticated architectures, more data, or more human feedback. Relying solely on LLMs for complex, high-stakes decisions is therefore problematic. Here we present a hybrid collective intelligence system that mitigates these risks by leveraging the complementary strengths of human experience and the vast information processed by LLMs. We apply our method to open-ended medical diagnostics, combining 40,762 differential diagnoses made by physicians with the diagnoses of five state-of-the art LLMs across 2,133 medical cases. We show that hybrid collectives of physicians and LLMs outperform both single physicians and physician collectives, as well as single LLMs and LLM ensembles. This result holds across a range of medical specialties and professional experience, and can be attributed to humans' and LLMs' complementary contributions that lead to different kinds of errors. Our approach highlights the potential for collective human and machine intelligence to improve accuracy in complex, open-ended domains like medical diagnostics.
△ Less
Submitted 21 June, 2024;
originally announced June 2024.
-
Animal behavior facilitates eco-evolutionary dynamics
Authors:
Kiyoko M. Gotanda,
Damien R. Farine,
Claudius F. Kratochwil,
Kate L. Laskowski,
Pierre-Olivier Montiglio
Abstract:
The mechanisms underlying eco-evolutionary dynamics (the feedback between ecological and evolutionary processes) are often unknown. Here, we propose that classical theory from behavioral ecology can provide a greater understanding of the mechanisms underlying eco-evolutionary dynamics, and thus improve predictions about the outcomes of these dynamics.
The mechanisms underlying eco-evolutionary dynamics (the feedback between ecological and evolutionary processes) are often unknown. Here, we propose that classical theory from behavioral ecology can provide a greater understanding of the mechanisms underlying eco-evolutionary dynamics, and thus improve predictions about the outcomes of these dynamics.
△ Less
Submitted 19 December, 2019;
originally announced December 2019.
-
Nested interaction networks represent a missing link in the study of behavioural and community ecology
Authors:
Pierre-Olivier Montiglio,
Kiyoko M. Gotanda,
Claudius F. Kratochwil,
Kate L. Laskowski,
Carey D. Nadell,
Damien R. Farine
Abstract:
Interactions are ubiquitous across biological systems. These interactions can be abstracted as patterns of connections among distinct units such as genes, proteins, individual organisms, or species which form a hierarchy of biological organisation. Connections in this hierarchy are arranged in a nested structure: gene and protein networks shape phenotypic traits and together constitute individuals…
▽ More
Interactions are ubiquitous across biological systems. These interactions can be abstracted as patterns of connections among distinct units such as genes, proteins, individual organisms, or species which form a hierarchy of biological organisation. Connections in this hierarchy are arranged in a nested structure: gene and protein networks shape phenotypic traits and together constitute individuals, individuals are embedded within populations, populations within communities, and communities within ecosystems. This pervasive "nestedness" of networks can result in propagation of the effects from within-level interactions at one level to units at higher or lower levels of organization. The concept of nested biological networks is implicit in a variety of disciplines ranging from the study of genetic circuits regulating phenotypic trait expression (Babu et al. 2004), to the study of predator-prey interactions influencing community composition (Poisot et al. 2016). However, studies typically only address interactions within and among directly neighbouring hierarchical levels, such as genotypes and phenotypic traits, or populations and communities. Here, we formalise nested networks as having nodes that can contain, or be embedded in, other nodes, and where edges can bridge connections between sets of embedded nodes. We then argue that explicitly accounting for network nestedness across levels of organization will encourage integrative thinking on new interdisciplinary research fronts. We focus on two phenomena in particular: (i) indirect connections among units can arise from the structure of connections at higher or lower levels of organisation, (ii) the propagation of effects across neighbouring hierarchical levels of organization. This framework of nested interaction networks provides a tool for researchers across disciplines to conceptualize their work as elements on a common scaffold.
△ Less
Submitted 17 September, 2018; v1 submitted 3 April, 2018;
originally announced April 2018.
-
Magnetic groundstate and Fermi surface of bcc Eu
Authors:
J. Kunes R. Laskowski
Abstract:
Using spin-spiral technique within the full potential linearized augmented-plane-waves (LAPW) electronic structure method we investigate the magnon spectrum and Néel temperature of bcc Eu. Ground state corresponding to an incommensurate spin-spiral is obtained in agreement with experiment and previous calculations. We demonstrate that the magnetic coupling is primarily through the intra-atomic…
▽ More
Using spin-spiral technique within the full potential linearized augmented-plane-waves (LAPW) electronic structure method we investigate the magnon spectrum and Néel temperature of bcc Eu. Ground state corresponding to an incommensurate spin-spiral is obtained in agreement with experiment and previous calculations. We demonstrate that the magnetic coupling is primarily through the intra-atomic $f-s$ and $f-d$ exchange and Ruderman-Kittel-Kasuya-Yosida mechanism. We show that the existence of this spin-spiral is closely connected to a nesting feature of the Fermi surface which was not noticed before.
△ Less
Submitted 15 April, 2004;
originally announced April 2004.
