Using the Veil of Ignorance to align AI systems with principles of justice.

The philosopher John Rawls proposed the Veil of Ignorance (VoI) as a thought experiment to identify fair principles for governing a society. Here, we apply the VoI to an important governance domain: artificial intelligence (AI). In five incentive-compatible studies (N = 2, 508), including two preregistered protocols, participants choose principles to govern an Artificial Intelligence (AI) assistant from behind the veil: that is, without knowledge of their own relative position in the group. Compared to participants who have this information, we find a consistent preference for a principle that instructs the AI assistant to prioritize the worst-off. Neither risk attitudes nor political preferences adequately explain these choices. Instead, they appear to be driven by elevated concerns about fairness: Without prompting, participants who reason behind the VoI more frequently explain their choice in terms of fairness, compared to those in the Control condition. Moreover, we find initial support for the ability of the VoI to elicit more robust preferences: In the studies presented here, the VoI increases the likelihood of participants continuing to endorse their initial choice in a subsequent round where they know how they will be affected by the AI intervention and have a self-interested motivation to change their mind. These results emerge in both a descriptive and an immersive game. Our findings suggest that the VoI may be a suitable mechanism for selecting distributive principles to govern AI.

Sex Differences in Systemic and Coronary Arterial Hemodynamics in Heart Failure With Preserved Ejection Fraction.

Heart failure (HF) with preserved ejection fraction (HFpEF) predominantly affects females. Systemic and coronary arterial abnormalities are present in HFpEF and may contribute to HFpEF in females. We performed a cross-sectional study of 32 participants with HFpEF and 26 controls. Arterial hemodynamics were noninvasively assessed by combining arterial tonometry with echocardiography. Coronary microvascular function was assessed by rubidium-82 positron emission tomography as the myocardial flow reserve. Coronary vascular resistance (CVR) at rest and vasodilator stress were calculated using positron emission tomography. CVR reserve was calculated as stress - rest CVR. Multivariable linear regression assessed the associations of female sex with arterial hemodynamics in participants with and without HF, and the association of HF with arterial hemodynamics within each sex stratum. Demographics and left ventricular systolic and diastolic function were similar between males and females. Among those with HFpEF, females had a higher steady and pulsatile arterial load and more impaired (less negative) CVR reserve than males. Conversely, in controls, females had similar hemodynamics to males. We then divided the sample based on sex. Femaleswith HFpEF had a higher pulsatile arterial load and higher stress CVR than control females. Among males, arterial hemodynamics were similar, regardless of HFpEF status. The measures of early pulsatile arterial load were independently associated with higher E/e' and lower myocardial flow reserve in females only. In conclusion, despite similar left ventricular function between sexes, older females with HFpEF are characterized by additional systemic and coronary arterial hemodynamic abnormalities compared with males with HFpEF and similarly aged females without HFpEF.

Combining Aortic Size With Arterial Hemodynamics Enhances Assessment of Future Thoracic Aortic Aneurysm Expansion.

BACKGROUND: Thoracic aortic aneurysm (TAA) is a deadly disease whose current method for risk stratification (aneurysm size) is imperfect. We sought to evaluate whether combining aortic size with hemodynamic measures that reflect the aorta's function was superior to aortic size alone in the assessment of TAA expansion. METHODS: One hundred thirty-seven nonoperated participants with TAA were followed prospectively. Aortic stiffness and pulsatile hemodynamics were noninvasively assessed at baseline with a combination of arterial tonometry with echocardiography using validated methodology. Aneurysm growth was calculated from standard imaging modalities. Multivariable linear regression models adjusted for potential confounders evaluated the association of aneurysm size and arterial hemodynamics, alone and in combination, with TAA growth. RESULTS: Sixty-nine percent of participants were male. Mean ± SD age, baseline aneurysm size, follow-up, and aneurysm expansion were, respectively, 62.2 ± 11.4 years, 45.9 ± 4.0 mm, 4.5 ± 1.9 years, and 0.41 ± 0.46 mm/year. In the linear regression models, the standardised ß (ß∗) for the association of aneurysm size with aneurysm expansion was 0.178 (P = 0.044). This was improved by combining aortic size with most measures of aortic function, with ß∗ ranging from 0.192 (for aneurysm size combined with central diastolic blood pressure) to 0.484 (for aneurysm size combined with carotid-femoral pulse-wave velocity) (P ≤ 0.05 for each). CONCLUSIONS: Combining aneurysm size with measures of arterial function improves assessment of aneurysm growth over TAA size alone, which is the standard for clinical decisions in TAA. Thus, combining aneurysm size with measures of aortic function provides a clinical advantage in the assessment of TAA disease activity.

Human-centred mechanism design with Democratic AI.

Building artificial intelligence (AI) that aligns with human values is an unsolved problem. Here we developed a human-in-the-loop research pipeline called Democratic AI, in which reinforcement learning is used to design a social mechanism that humans prefer by majority. A large group of humans played an online investment game that involved deciding whether to keep a monetary endowment or to share it with others for collective benefit. Shared revenue was returned to players under two different redistribution mechanisms, one designed by the AI and the other by humans. The AI discovered a mechanism that redressed initial wealth imbalance, sanctioned free riders and successfully won the majority vote. By optimizing for human preferences, Democratic AI offers a proof of concept for value-aligned policy innovation.

PDCD10/CCM3 acts downstream of {gamma}-protocadherins to regulate neuronal survival.

γ-Protocadherins (PCDH-γ) regulate neuronal survival in the vertebrate central nervous system. The molecular mechanisms of how PCDH-γ mediates this function are still not understood. In this study, we show that through their common cytoplasmic domain, different PCDH-γ isoforms interact with an intracellular adaptor protein named PDCD10 (programmed cell death 10). PDCD10 is also known as CCM3, a causative genetic defect for cerebral cavernous malformations in humans. Using RNAi-mediated knockdown, we demonstrate that PDCD10 is required for the occurrence of apoptosis upon PCDH-γ depletion in developing chicken spinal neurons. Moreover, overexpression of PDCD10 is sufficient to induce neuronal apoptosis. Taken together, our data reveal a novel function for PDCD10/CCM3, acting as a critical regulator of neuronal survival during development.

Multiple Spontaneous Coronary Artery Dissections: An Uncommon Cause of Acute Coronary Syndrome in a Syrian Refugee.

Spontaneous coronary artery dissection is an uncommon nonatherosclerotic cause of acute coronary syndrome. It usually occurs in young women and is often associated with fibromuscular dysplasia, connective tissue diseases, and pregnancy or postpartum states. We present a case of a Syrian woman with a history of grand multiparity and recent miscarriage who presented with non-ST-elevation myocardial infarction and was found to have multivessel spontaneous coronary artery dissection and severe left ventricular dysfunction.

Clinical presentation and outcome of patients with ST-segment elevation myocardial infarction without culprit angiographic lesions.

Patients diagnosed with ST-segment elevation myocardial infarction (STEMI) are occasionally found to have no culprit lesion on coronary angiography and are classified as presenting with false-positive STEMI. The clinical presentation and outcomes of these patients need to be further explored. In this case-controlled study, 259 consecutive patients with true code STEMI were compared to 104 consecutive STEMI patients without culprit lesions on emergent coronary angiography. We compared the clinical presentation, electrocardiographic features, etiology, and outcomes of the two groups. STEMI patients without culprit lesions were less likely to have typical chest pain (46% vs. 79%, P < 0.01). The ST-elevation in the group without culprit lesion was more likely to be concave (56% vs. 31%, P < 0.01), with less reciprocal ST-depression (19% vs. 71%, P < 0.01). The group without culprit lesions had a higher rate of ventilator support requirement (12.4% vs. 5.4%, P = 0.02), and higher rate of 30-day mortality (11.0% vs. 5.9%, P = 0.02). However, after excluding the patients with out-of-hospital cardiac arrests from both groups, the difference was no longer significant (P = 0.40 and 0.34 respectively). The relative poor outcomes of patients with false-positive code STEMI reflect the severity of their underlying medical condition. Careful history and review of ECG may help differentiate this group from true STEMI.

Diagnostic accuracy of ST-segment elevation myocardial infarction by various healthcare providers.

BACKGROUND: This study aimed to compare the accuracy of ECG interpretation for diagnosis of STEMI by different groups of healthcare professionals involved in the STEMI program at our institution. METHODS: We selected 21 ECGs from patients with typical symptoms of MI that were diagnosed with STEMI, and 10 ECGs of STEMI mimics. STEMI mimic ECGs were repeated in the package with a story of typical and atypical chest pain. ECGs were interpreted to diagnose STEMI and identify need for initiation of the cardiac catheterization lab (CCL). Participants identified confidence in STEMI recognition, and average number of ECGs read per week. RESULTS: A total of 64 participants completed the study package. Cardiologists were more likely to provide correct interpretation compared to other groups. False positive diagnoses were more likely made by paramedics when compared to cardiologists (p < 0.01). There was a positive correlation between increased exposure to ECGs and accurate STEMI diagnosis (r = 0.482, p < 0.001). A threshold of ≥ 20 ECGs read per week showed a statistically significant improvement in accuracy (p < 0.001). Self-reported confidence correlated positively with accuracy (r = 0.402, p =< 0.001). Changing the ECG narrative of the STEMI mimic ECGs had a significant effect on interpretation between groups (p = 0.043). CONCLUSIONS: Our study showed that healthcare profession and number of ECGs reviewed per week are predictive of the accuracy of ECG interpretation of STEMI. Cardiologists are the most accurate diagnosticians, and are the least likely to falsely activate the CCL. Weekly exposure of ≥ 20 ECGs may improve diagnostic accuracy regardless of underlying experience.

Characterization of fibrinogen-like protein 2 (FGL2): monomeric FGL2 has enhanced immunosuppressive activity in comparison to oligomeric FGL2.

Fibrinogen-like protein 2 (FGL2), a novel effector molecule of CD4(+)CD25(+)Foxp3(+) regulatory T cells (Treg), mediates its suppressive activity through binding to low affinity Fcγ receptors expressed on antigen presenting cells (APCs). FGL2 has been implicated in the pathogenesis of viral hepatitis, xeno- and allotransplant rejection, and rheumatoid arthritis. Here we fully analyzed the structure-function relationships of recombinant murine FGL2 generated in COS-7 cells and identified the receptor binding domains. Native FGL2 exists as an oligomer with a molecular weight of approximately 260 kDa, while under reducing conditions, FGL2 has a molecular weight of 65 kDa suggesting that native FGL2 is composed of four monomers. By site-directed mutation, cysteines at positions 94, 97, 184 and 187, found in the coiled-coil domain were shown to be crucial for FGL2 oligomerization. Monomeric FGL2 had a lower affinity binding to APCs, but increased immunosuppressive activity compared to oligomeric FGL2. Deglycosylation demonstrated that sugar moieties are critical for maintaining solubility of FGL2. SWISS-MODEL analysis suggested that FGL2 has a similar tertiary structure with other members of the fibrinogen family such as fibrinogen and tachylectin. Mutational analysis of cysteine residues and Western blots suggested an asymmetric bouquet-shaped quaternary structure for oligomeric FGL2, resembling many pattern-recognition molecules in the lectin pathway of innate immunity. The functional motifs of FGL2 were mapped to the C terminal globular domain, using a peptide blockade assay. These results collectively define the biochemical and immunological determinants of FGL2, an important immunosuppressive molecule of Treg providing important insights for designing FGL2-related therapeutics.

Admixture analysis of age at onset in major depressive disorder.

OBJECTIVES: This study aimed to determine the distributions of the age at onset (AAO) in patients with major depressive disorder (MDD) using admixture analysis and to determine the clinical differences between subgroups with different AAO. METHODS: Participants were administered the Mini-International Neuropsychiatric Interview to obtain clinical data. Admixture analysis was performed using the STATA module DENORMIX to identify subgroups characterized by differences in AAO. RESULTS: The best fit model was the three-component model with the following means, standard deviations and proportions: 14.60 (3.75) years (49.1%), 29.15 (6.75) years (34.1%) and 46.96 (6.06) years (16.8%) (χ(2)=3.64, 2 df, P=.162). The three subgroups were divided by AAO of 22 and 40. After controlling for duration of illness, there were no significant differences between the three AAO subgroups in terms of gender and psychiatric family history. However, the early-onset subgroup was significantly more likely to report being single compared to the intermediate- and late-onset groups. The proportion of individuals meeting criteria for lifetime comorbid panic disorders and obsessive-compulsive disorder did not differ significantly between the AAO groups. However, the early-onset group reported a higher incidence of attention-deficit/hyperactivity disorder (5.1% vs. 1.7% and 1.2%, P=.086), although this was not statistically significant. CONCLUSIONS: Our study identified three characteristically different AAO subgroups in individuals suffering from MDD. The subgroups may reflect different underlying neurobiological mechanisms involved.