Online images amplify gender bias.

Each year, people spend less time reading and more time viewing images1, which are proliferating online2-4. Images from platforms such as Google and Wikipedia are downloaded by millions every day2,5,6, and millions more are interacting through social media, such as Instagram and TikTok, that primarily consist of exchanging visual content. In parallel, news agencies and digital advertisers are increasingly capturing attention online through the use of images7,8, which people process more quickly, implicitly and memorably than text9-12. Here we show that the rise of images online significantly exacerbates gender bias, both in its statistical prevalence and its psychological impact. We examine the gender associations of 3,495 social categories (such as 'nurse' or 'banker') in more than one million images from Google, Wikipedia and Internet Movie Database (IMDb), and in billions of words from these platforms. We find that gender bias is consistently more prevalent in images than text for both female- and male-typed categories. We also show that the documented underrepresentation of women online13-18 is substantially worse in images than in text, public opinion and US census data. Finally, we conducted a nationally representative, preregistered experiment that shows that googling for images rather than textual descriptions of occupations amplifies gender bias in participants' beliefs. Addressing the societal effect of this large-scale shift towards visual communication will be essential for developing a fair and inclusive future for the internet.

Divergences in color perception between deep neural networks and humans.

Deep neural networks (DNNs) are increasingly proposed as models of human vision, bolstered by their impressive performance on image classification and object recognition tasks. Yet, the extent to which DNNs capture fundamental aspects of human vision such as color perception remains unclear. Here, we develop novel experiments for evaluating the perceptual coherence of color embeddings in DNNs, and we assess how well these algorithms predict human color similarity judgments collected via an online survey. We find that state-of-the-art DNN architectures - including convolutional neural networks and vision transformers - provide color similarity judgments that strikingly diverge from human color judgments of (i) images with controlled color properties, (ii) images generated from online searches, and (iii) real-world images from the canonical CIFAR-10 dataset. We compare DNN performance against an interpretable and cognitively plausible model of color perception based on wavelet decomposition, inspired by foundational theories in computational neuroscience. While one deep learning model - a convolutional DNN trained on a style transfer task - captures some aspects of human color perception, our wavelet algorithm provides more coherent color embeddings that better predict human color judgments compared to all DNNs we examine. These results hold when altering the high-level visual task used to train similar DNN architectures (e.g., image classification versus image segmentation), as well as when examining the color embeddings of different layers in a given DNN architecture. These findings break new ground in the effort to analyze the perceptual representations of machine learning algorithms and to improve their ability to serve as cognitively plausible models of human vision. Implications for machine learning, human perception, and embodied cognition are discussed.

Time-Dependent Changes in THC Concentrations in Deceased Persons.

Changes in the concentrations of Δ9-tetrahydrocannabinol (THC) in the postmortem period were investigated in a series of cases by comparing concentrations in blood taken on receipt of the body in the mortuary (admission specimen, AD) with the concentrations obtained in blood taken at autopsy some time later and also from blood specimens taken antemortem. Overall, the median THC concentration in AD blood was 13.7 ng/mL (n = 239, range LOQ-220), while the median concentration at autopsy was 13.8 ng/mL (n = 106, range LOQ-810) and 1.9 ng/mL (n = 147, range LOQ-48) antemortem. Fourteen cases had all three specimens taken from the same decedent. The corresponding AM, AD and PM median concentrations were 4.0 (range LOQ-48), 15.5 (range 4.0-176) and 4.4 ng/mL (LOQ-56), respectively. The median elapsed times from AM to AD and AD to PM were 33 and 97.5 h, respectively. In contrast, acetaminophen showed no change in blood concentration from AM to AD (6.8 and 6.0 mg/L, respectively). These data show large increases in THC concentration in the early postmortem period, followed by a decline, although the median blood concentrations at autopsy were similar to that obtained antemortem. In contrast, when blood was taken from the femoral region, subclavian and heart ventricles sites, in the same case, the THC concentrations, while variable, showed overall no significant difference. These dynamic changes reflect complex phenomenon occurring in deceased persons and will further serve to increase the uncertainty over any interpretation of postmortem THC concentrations.

Color associations in abstract semantic domains.

The embodied cognition paradigm has stimulated ongoing debate about whether sensory data - including color - contributes to the semantic structure of abstract concepts. Recent uses of linguistic data in the study of embodied cognition have been focused on textual corpora, which largely precludes the direct analysis of sensory information. Here, we develop an automated approach to multimodal content analysis that detects associations between words based on the color distributions of their Google Image search results. Crucially, we measure color using a transformation of colorspace that closely resembles human color perception. We find that words in the abstract domains of academic disciplines, emotions, and music genres, cluster in a statistically significant fashion according to their color distributions. Furthermore, we use the lexical ontology WordNet and crowdsourced human judgments to show that this clustering reflects non-arbitrary semantic structure, consistent with metaphor-based accounts of embodied cognition. In particular, we find that images corresponding to more abstract words exhibit higher variability in colorspace, and semantically similar words have more similar color distributions. Strikingly, we show that color associations often reflect shared affective dimensions between abstract domains, thus revealing patterns of aesthetic coherence in everyday language. We argue that these findings provide a novel way to synthesize metaphor-based and affect-based accounts of embodied semantics.

Ultra-rapid targeted analysis of 40 drugs of abuse in oral fluid by LC-MS/MS using carbon-13 isotopes of methamphetamine and MDMA to reduce detector saturation.

The number of oral fluid samples collected by the road policing authority in Victoria, Australia, requiring confirmatory laboratory analysis for drugs proscribed under Victorian legislation (methamphetamine, MDMA and Δ9-tetrahydrocannabinol) has greatly increased in recent years, driving the need for improved analysis techniques to enable expedient results. The aim of this study was to develop an LC-MS/MS-based targeted oral fluid screening technique that covers a broad range of basic and neutral drugs of abuse that can satisfy increased caseload while monitoring other compounds of interest for epidemiological purposes. By combining small sample volume, simple extraction procedure, rapid LC-MS/MS analysis and automated data processing, 40 drugs of abuse including amphetamines, benzodiazepines, cocaine and major metabolites, opioids, cannabinoids and some designer stimulants were separated over 5 min (with an additional 0.5 min re-equilibration time). The analytes were detected using a Sciex® API 4500 Q-Trap LC-MS/MS system with positive ESI in MRM mode monitoring three transitions per analyte. The method was fully validated in accordance with international guidelines and also monitored carbon-13 isotopes of MDMA and MA to reduce detector saturation effects, allowing for confirmation of large concentrations of these compounds without the need for dilution or re-analysis. The described assay has been successfully used for analysis of oral fluid collected as part of law enforcement procedures at the roadside in Victoria, providing forensic results as well as epidemiological prevalence in the population tested. The fast and reliable detection of a broad range of drugs and subsequent automated data processing gives the opportunity for high throughput and fast turnaround times for forensic toxicology.

Residual cannabis levels in blood, urine and oral fluid following heavy cannabis use.

An understanding of tetrahydrocannabinol (THC) kinetics and residual levels after cannabis use is essential in interpreting toxicology tests in body fluids from live subjects, particularly when used in forensic settings for drug abuse, traffic and interpersonal violence cases. However the current literature is largely based on laboratory studies using controlled cannabis dosages in experienced users, with limited research investigating the kinetics of residual THC concentrations in regular high dose cannabis users. Twenty-one dependent cannabis users were recruited at admission to two residential detoxification units in Melbourne, Australia. After being provided with information about, and consenting to, the study, subjects volunteered to provide once-daily blood, urine and oral fluid (saliva) samples for seven consecutive days following admission, involving cessation and abstinence from all cannabis use. Blood and oral fluid specimens were analysed for THC and urine specimens for the metabolite THC-COOH. In some subjects THC was detectable in blood for at least 7 days and oral fluid specimens were positive for THC up to 78 h after admission to the unit. Urinary THC-COOH concentrations exceeded 1000 ng/mL for some subjects 129 h after last use. The presented blood THC levels are higher and persist longer in some individuals than previously described, our understanding and interpretation of THC levels in long term heavy cannabis users may need to be reconsidered.

The incidence of drugs of impairment in oral fluid from random roadside testing.

Oral fluid (OF) has become a popular specimen to test for presence of drugs, particularly in regards to road safety. In Victoria, OF specimens from drivers have been used to test for the presence of methylamphetamine (MA) and Δ(9)-tetrahydrocannabinol (THC) since 2003 and 3,4-methylenedioxy-N-methylamphetamine (MDMA) since 2006. LC-MS/MS has been used to test the most recent 853 submitted OF specimens from Victoria Police for 31 drugs of abuse including those listed in the Australian Standard AS4760-2006. At least one proscribed drug was detected in 96% of drivers, of which MA was the most common (77%), followed by THC (42%), MDMA (17%) and the combination of all three (3.9%). Opioids were detected in 14% of drivers of which 4.8% were positive for 6-acetylmorphine and 3.3% for methadone. The incidence of the opioids tramadol (1.2%) and oxycodone (1.1%) were relatively low. Cocaine (8.0%) was as commonly detected as benzodiazepines (8.0%), and was almost always found in combination with MA (7.9%). Samples positive to benzodiazepines were largely due to diazepam (3.5%) and alprazolam (3.4%), with only 0.2% of drivers combining the two. Ketamine was also detected in 1.5% of cases. While the incidences of the proscribed drugs itself are concerning, it is clear that many drivers are also using other drugs capable of causing impairment.

Drugs in oral fluid in randomly selected drivers.

There were 13,176 roadside drug tests performed in the first year of the random drug-testing program conducted in the state of Victoria. Drugs targeted in the testing were methamphetamines and Delta(9)-tetrahydrocannabinol (THC). On-site screening was conducted by the police using DrugWipe, while the driver was still in the vehicle and if positive, a second test on collected oral fluid, using the Rapiscan, was performed in a specially outfitted "drug bus" located adjacent to the testing area. Oral fluid on presumptive positive cases was sent to the laboratory for confirmation with limits of quantification of 5, 5, and 2 ng/mL for methamphetamine (MA), methylenedioxy-methamphetamine (MDMA), and THC, respectively. Recovery experiments conducted in the laboratory showed quantitative recovery of analytes from the collector. When oral fluid could not be collected, blood was taken from the driver and sent to the laboratory for confirmation. These roadside tests gave 313 positive cases following GC-MS confirmation. These comprised 269, 118, and 87 cases positive to MA, MDMA, and THC, respectively. The median oral concentrations (undiluted) of MA, MDMA, and THC was 1136, 2724, and 81 ng/mL. The overall drug positive rate was 2.4% of the screened population. This rate was highest in drivers of cars (2.8%). The average age of drivers detected with a positive drug reading was 28 years. Large vehicle (trucks over 4.5 t) drivers were older; on average at 38 years. Females accounted for 19% of all positives, although none of the positive truck drivers were female. There was one false positive to cannabis when the results of both on-site devices were considered and four to methamphetamines.

Phase-referenced probe interferometer for biological surface profiling and displacement measurements.

We present a probe-based, phase-referenced low coherence interferometer in which the reference field is provided by a fiber end reflection. A gradient-index microlens focuses light onto a sample and collects reflected light. We use the probe interferometer to measure surface profiles of the compound eye of a housefly (Musca domestica) and measure nanometer-scale vibrations in a test sample.

Noncontact measurement of nerve displacement during action potential with a dual-beam low-coherence interferometer.

We have used a novel phase-referenced heterodyne dual-beam low-coherence interferometer to perform what we believe are the first noncontact measurements of surface motion in a nerve bundle during the action potential. Nerve displacements of approximately 5-nm amplitude and approximately 10-ms duration are measured without signal averaging. This interferometer may find general application in measurement of small motion in cells and other weakly scattering samples.

The involvement of drugs in drivers of motor vehicles killed in Australian road traffic crashes.

A multi-center case-control study was conducted on 3398 fatally-injured drivers to assess the effect of alcohol and drug use on the likelihood of them being culpable. Crashes investigated were from three Australian states (Victoria, New South Wales and Western Australia). The control group of drug- and alcohol-free drivers comprised 50.1% of the study population. A previously validated method of responsibility analysis was used to classify drivers as either culpable or non-culpable. Cases in which the driver "contributed" to the crash (n=188) were excluded. Logistic regression was used to examine the association of key attributes such as age, gender, type of crash and drug use on the likelihood of culpability. Drivers positive to psychotropic drugs were significantly more likely to be culpable than drug-free drivers. Drivers with Delta(9)-tetrahydrocannabinol (THC) in their blood had a significantly higher likelihood of being culpable than drug-free drivers (odds ratio (OR) 2.7, 95% CI 1.02-7.0). For drivers with blood THC concentrations of 5 ng/ml or higher the odds ratio was greater and more statistically significant (OR 6.6, 95% CI 1.5-28.0). The estimated odds ratio is greater than that for drivers with a blood alcohol concentration (BAC) of 0.10-0.15% (OR 3.7, 95% CI 1.5-9.1). A significantly stronger positive association with culpability was seen with drivers positive to THC and with BAC > or =0.05% compared with BAC > or =0.05 alone (OR 2.9, 95% CI 1.1-7.7). Strong associations were also seen for stimulants, particularly in truck drivers. There were non-significant, weakly positive associations of opiates and benzodiazepines with culpability. Drivers positive to any psychoactive drug were significantly more likely to be culpable (OR 1.8, 95% CI 1.3-2.4). Gender differences were not significant, but differences were apparent with age. Drivers showing the highest culpability rates were in the under 25 and over 65 age groups.

The incidence of drugs in drivers killed in Australian road traffic crashes.

The incidence of alcohol and drugs in fatally injured drivers were determined in three Australian states; Victoria (VIC), New South Wales (NSW) and Western Australia (WA) for the period of 1990-1999. A total of 3398 driver fatalities were investigated which included 2609 car drivers, 650 motorcyclists and 139 truck drivers. Alcohol at or over 0.05 g/100ml (%) was present in 29.1% of all drivers. The highest prevalence was in car drivers (30.3%) and the lowest in truckers (8.6%). WA had the highest rate of alcohol presence of the three states (35.8%). Almost 10% of the cases involved both alcohol and drugs. Drugs (other than alcohol) were present in 26.7% of cases and psychotropic drugs in 23.5%. These drugs comprised cannabis (13.5%), opioids (4.9%), stimulants (4.1%), benzodiazepines (4.1%) and other psychotropic drugs (2.7%). 8.5% of all drivers tested positive for Delta(9)-tetrahydrocannabinol (THC) and the balance of cannabis positive drivers were positive to only the 11-nor-Delta(9)-tetrahydrocannabinol-9-carboxylic acid (carboxy-THC) metabolite. The range of THC blood concentrations in drivers was 0.1-228 ng/ml, with a median of 9 ng/ml. Opioids consisted mainly of morphine (n=84), codeine (n=89) and methadone (n=33), while stimulants consisted mainly of methamphetamine (n=51), MDMA (n=6), cocaine (n=5), and the ephedrines (n=61). The prevalence of drugs increased over the decade, particularly cannabis and opioids, while alcohol decreased. Cannabis had a larger prevalence in motorcyclists (22.2%), whereas stimulants had a much larger presence in truckers (23%).

Determination of delta9-THC in whole blood using gas chromatography-mass spectrometry.

A simple and reliable liquid-liquid extraction method for the determination of delta9-tetrahydrocannabinol (THC) in whole blood utilizing gas chromatography-mass spectrometry in electron impact mode is described. The substance is derivatized with pentafluoropropionic anhydride in pentafluropropanol. The limit of detection is 0.5 ng/mL for a 1-mL specimen, with recovery greater than 70%. The intra-assay coefficient of variation (CV) is 3.1% to 5.2%, and the interassay CV is 6.4% to 9.5%, calculated at THC concentrations of 1, 5, and 25 ng/mL. The accuracy is between 95 and 97%. The optimization of extraction and derivatization conditions is detailed.