The increasingly pervasive role of Artificial Intelligence (AI) in our societies is radically changing the way that social interaction takes place within all fields of knowledge. The obvious ...opportunities in terms of accuracy, speed and originality of research are accompanied by questions about the possible risks and the consequent responsibilities involved in such a disruptive technology. In recent years, this twofold aspect has led to an increase in analyses of the ethical and political implications of AI. As a result, there has been a proliferation of documents that seek to define the strategic objectives of AI together with the ethical precautions required for its acceptable development and deployment. Although the number of documents is certainly significant, doubts remain as to whether they can effectively play a role in safeguarding democratic decision-making processes. Indeed, a common feature of the national strategies and ethical guidelines published in recent years is that they only timidly address how to integrate civil society into the selection of AI objectives. Although scholars are increasingly advocating the necessity to include civil society, it remains unclear which modalities should be selected. If both national strategies and ethics guidelines appear to be neglecting the necessary role of a democratic scrutiny for identifying challenges, objectives, strategies and the appropriate regulatory measures that such a disruptive technology should undergo, the question is then, what measures can we advocate that are able to overcome such limitations? Considering the necessity to operate holistically with AI as a social object, what theoretical framework can we adopt in order to implement a model of governance? What conceptual methodology shall we develop that is able to offer fruitful insights to governance of AI? Drawing on the insights of classical pragmatist scholars, we propose a framework of democratic experimentation based on the method of social inquiry. In this article, we first summarize some of the main points of discussion around the potential societal, ethical and political issues of AI systems. We then identify the main answers and solutions by analyzing current national strategies and ethics guidelines. After showing the theoretical and practical limits of these approaches, we outline an alternative proposal that can help strengthening the active role of society in the discussion about the role and extent of AI systems.
Background. Oseltamivir provides modest clinical benefits to children with influenza when started within 48 hours of symptom onset. The effectiveness of oseltamivir could be substantially greater if ...the treatment were started earlier during the course of the illness. Methods. We carried out a randomized, double-blind, placebo-controlled trial of the efficacy of oseltamivir started within 24 hours of symptom onset in children 1–3 years of age with laboratory-confirmed influenza during the seasons of 2007–2008 and 2008–2009. Eligible children received either orally administered oseltamivir suspension or a matching placebo twice daily for 5 days. The children received clinical examinations, and the parents filled out detailed symptom diaries for 21 days. Results. Of 408 randomized children who received the study drug (oseltamivir, 203, and placebo, 205), 98 had laboratory-confirmed influenza (influenza A, 79, and influenza B, 19). When started within 12 hours of the onset of symptoms, oseltamivir decreased the incidence of acute otitis media by 85% (95% confidence interval, 25%–97%), but no significant reduction was observed with treatment started within 24 hours. Among children with influenza A, oseltamivir treatment started within 24 hours shortened the median time to resolution of illness by 3.5 days (3.0 vs 6.5 days; P = .002) in all children and by 4.0 days (3.4 vs 7.3; P = .006) in unvaccinated children and reduced parental work absenteeism by 3.0 days. No efficacy was demonstrated against influenza B infections. Conclusions. Oseltamivir treatment started within 24 hours of symptom onset provides substantial benefits to children with influenza A infection. Clinical trials registration. ClinicalTrials.gov identifier: NCT00593502.
Summary Background Few prospectively collected data are available to support the effectiveness of inactivated influenza vaccines in children younger than 2 years. We aimed to establish the ...effectiveness of trivalent inactivated influenza vaccine against laboratory-confirmed influenza A and B infections in a cohort of children younger than 3 years. Methods In a prospective cohort study during the influenza season of 2007–08 in Turku, Finland, between Jan 14 and April 9, 2008, we assessed the effectiveness of trivalent inactivated influenza vaccine against laboratory-confirmed influenza A and B infections in children aged 9 months to 3 years. Our study was part of a clinical trial on antiviral treatment of influenza in children ( ClinicalTrials.gov , number NCT00593502 ). The vaccine was given as part of the Finnish vaccination programme as a 0·5 mL injection. Children enrolled into our study through mailed announcements and local advertisements were examined every time they had fever or signs of respiratory infection. No exclusion criteria were used for enrolment. Influenza was diagnosed with viral culture, antigen detection, and RT-PCR assays of nasal swab specimens. Vaccination status of children was determined by parental report. We calculated the primary effectiveness of influenza vaccination by comparing the proportions of infections in fully vaccinated and unvaccinated children in the follow-up cohort. Findings We enrolled 631 children into our study with a mean age of 2·13 years (range 9–40 months). Seven (5%) of 154 fully vaccinated children and 61 (13%) of 456 unvaccinated children contracted influenza during the study (effectiveness 66%, 95% CI 29–84; p=0·003). In the subgroup of children younger than 2 years, four (4%) of 96 fully vaccinated children and 21 (12%) of 172 unvaccinated children contracted influenza (66%, 9–88, p=0·03). We were unable to record any adverse events associated with the vaccination of the children in our study. Interpretation Trivalent inactivated influenza vaccine was effective in preventing influenza in young children, including those younger than 2 years. Our findings suggest that influenza vaccine recommendations should be reassessed in most countries. Funding F Hoffmann-La Roche Ltd.
Industrial processes, coal combustion, biomass burning (BB), and vehicular transport are important sources of atmospheric fine particles (PM2.5) and contribute to ambient air concentrations of ...health-hazardous species, such as heavy metals, polycyclic aromatic hydrocarbons (PAH), and oxygenated-PAHs (OPAH). In China, emission controls have been implemented to improve air quality during large events, like the Youth Olympic Games (YOG) in August 2014 in Nanjing. In this work, six measurement campaigns between January 2014 and August 2015 were undertaken in Nanjing to determine the effects of emission controls and meteorological factors on PM2.5 concentration and composition. PAHs, OPAHs, hopanes, n‑alkanes, heavy metals, and several other inorganic elements were measured. PM2.5 and potassium concentrations were the highest in May–June 2014 indicating the prevalence of BB plumes in Nanjing. Emission controls substantially reduced concentrations of PM2.5 (31%), total PAHs (59%), OPAHs (37%), and most heavy metals (44–89%) during the YOG compared to August 2015. In addition, regional atmospheric transport and meteorological parameters partly explained the observed differences between the campaigns. The most abundant PAHs and OPAHs were benzob,kfluoranthenes, fluoranthene, pyrene, chrysene, 1,8‑naphthalic anhydride, and 9,10‑anthracenedione in all campaigns. Carbon preference index and the contribution of wax n‑alkanes indicated mainly biogenic sources of n‑alkanes in May–June 2014 and anthropogenic sources in the other campaigns. Hopane indexes pointed to vehicular transport as the major source of hopanes, but contribution of coal combustion was detected in winter 2015. The results provide evidence to the local government of the impacts of the air protection regulations. However, differences between individual components were observed, e.g., concentrations of potentially more harmful OPAHs decreased less than concentrations of PAHs. The results suggest that the proportions of hazardous components in the PM2.5 may also change considerably due to emission control measures.
Display omitted
•PM2.5 composition was analyzed before, during, and after the 2014 YOG in Nanjing.•Emission controls reduced concentrations of PAHs, OPAHs, and most heavy metals.•Different reduction rates between individual components were observed.•Specific chemical markers showed seasonal variation in the emission sources.
Publications in atmospheric sciences typically neglect biases caused by regression dilution (bias of the ordinary least squares line fitting) and regression to the mean (RTM) in comparisons of ...uncertain data. We use synthetic observations mimicking real atmospheric data to demonstrate how the biases arise from random data uncertainties of measurements, model output, or satellite retrieval products. Further, we provide examples of typical methods of data comparisons that have a tendency to pronounce the biases. The results show, that data uncertainties can significantly bias data comparisons due to regression dilution and RTM, a fact that is known in statistics but disregarded in atmospheric sciences. Thus, we argue that often these biases are widely regarded as measurement or modeling errors, for instance, while they in fact are artificial. It is essential that atmospheric and geoscience communities become aware of and consider these features in research.
Key Points
Regression to the mean and regression dilution can cause artificial bias in comparisons of unbiased data
We demonstrate how these effects are stronger when data uncertainty is larger
These effects are widely neglected in atmospheric sciences, potentially causing bias in many published results
The effects of aerosol on cloud droplet effective radius (R
eff
), cloud optical thickness and cloud droplet number concentration (N
d
) are analysed both from long-term direct ground-based in situ ...measurements conducted at the Puijo measurement station in Eastern Finland and from the Moderate Resolution Imaging Spectroradiometer (MODIS) instrument onboard the Terra and Aqua satellites. The mean in situ N
d
during the period of study was 217 cm
−3
, while the MODIS-based N
d
was 171 cm
−3
. The absolute values, and the dependence of both N
d
observations on the measured aerosol number concentration in the accumulation mode (N
acc
), are quite similar. In both data sets N
d
is clearly dependent on N
acc
, for N
acc
values lower than approximately 450 cm
−3
. Also, the values of the aerosol-cloud-interaction parameter ACI=(1/3)*d ln(N
d
)/d ln(N
acc
) are quite similar for N
acc
<400 cm
−3
with values of 0.16 and 0.14 from in situ and MODIS measurements, respectively. With higher N
acc
(>450 cm
−3
) N
d
increases only slowly. Similarly, the effect of aerosol on MODIS-retrieved R
eff
is visible only at low N
acc
values. In a sub set of data, the cloud and aerosol properties were measured simultaneously. For that data the comparison between MODIS-derived N
d
and directly measured N
d
, or the cloud droplet number concentration estimated from N
acc
values (N
d,p
), shows a correlation, which is greatly improved after careful screening using a ceilometer to make sure that only single cloud layers existed. This suggests that such determination of the number of cloud layers is very important when trying to match ground-based measurements to MODIS measurements.