Sunday 20 August 2017

New Techniques To Measure Social Bias in Software

Today, banks are increasingly using software to decide who will get a loan, courts to judge who should be denied bail, and hospitals to choose treatments for patients. These uses of software make it critical that the software does not discriminate against groups or individuals, say computer science researchers at the University of Massachusetts Amherst.
Professor Alexandra Meliou in the College of Information and Computer Sciences says, "The increased role of software and the potential impact it has on people's lives makes software fairness a critical property. Data-driven software has the ability to shape human behavior: it affects the products we view and purchase, the news articles we read, the social interactions we engage in, and, ultimately, the opinions we form."
Meliou with professor Yuriy Brun and Ph.D. student Sainyam Galhotra, have developed a new technique they call "Themis," to automatically test software for discrimination. They hope Themis will empower stakeholders to better understand software behavior, judge when unwanted bias is present, and, ultimately improve the software.
Brun says, "Unchecked, biases in data and software run the risk of perpetuating biases in society. For example, prior work has demonstrated that racial bias exists in online advertising delivery systems, where online searches for traditionally-minority names were more likely to yield ads related to arrest records. Such software behavior can contribute to racial stereotypes and other grave societal consequences."
The researchers' paper describing this research, published in pre-conference materials for the European Software Engineering Conference (ESEC/FSE 2017) before its September meeting in Paderborn, Germany, has won an Association of Computing Machinery Special Interest Group on Software Engineering (ACM SIGSOFT) Distinguished Paper Award. The work is supported by the National Science Foundation.
Brun explains that while earlier research has considered discrimination in software, Themis focuses on measuring causality in discrimination. Software testing allows Themis to perform hypothesis testing, to ask such questions as whether changing a person's race affects whether the software recommends giving that person a loan, he says.
"Our approach measures discrimination more accurately than prior work that focused on identifying differences in software output distributions, correlations or mutual information between inputs and outputs. Themis can identify bias in software whether that bias is intentional or unintentional, and can be applied to software that relies on machine learning, which can inject biases from data without the developers' knowledge," he adds.
When evaluated on public software systems from GitHub, Themis found that discrimination can sneak in even when the software is explicitly designed to be fair. State-of-the-art techniques for removing discrimination from algorithms fail in many situations, in part because prior definitions of discrimination failed to capture causality, the researchers point out.
For example, Themis found that a decision-tree-based machine learning approach specifically designed not to discriminate against gender was actually discriminating more than 11 percent of the time. That is, more than 11 percent of the individuals saw the software output affected just by altering their gender.
Themis also found that designing the software to avoid discrimination against one attribute may increase discrimination against others. For example, the same decision-tree-based software trained not to discriminate on gender discriminated against race 38 percent of the time. "These systems learn discrimination from biased data, but without careful control for potential bias, software can magnify that bias even further," Galhotra says.
Source: sciencedaily.com

Friday 4 August 2017

Insufficient Sleep May Be Adding To Your Waistline

Adults in the UK who have poor sleep patterns are more likely to be overweight and obese and have poorer metabolic health, according to a new study.
The findings showed that people who were sleeping on average around six hours a night had a waist measurement that was 3cm greater than individuals who were getting nine hours of sleep a night. And shorter sleepers were heavier too.
The results strengthen the evidence that insufficient sleep could contribute to the development of metabolic diseases such as diabetes -- major health challenges facing the NHS.
The study -- led by Dr Laura Hardie, Reader in Molecular Epidemiology at the University of Leeds -- not only looked at the links between sleep duration, diet and weight, but also other indicators of overall metabolic health such as blood pressure, blood cholesterol, blood sugar, and thyroid function.
The study involved 1,615 adults who reported how long they slept and kept records of food intake. Participants had blood samples taken and their weight, waist circumference, and blood pressure recorded.
The researchers looked at the associations between how long people were sleeping and these key biological parameters.
The research team, from the Leeds Institute of Cardiovascular and Metabolic Medicine and the School of Food Science and Nutrition, reported their findings in the journal PLOS One.
Obesity has doubled
Greg Potter, one of the Leeds researchers, said "The number of people with obesity worldwide has more than doubled since 1980.
"Obesity contributes to the development of many diseases, most notably type 2 diabetes. Understanding why people gain weight has crucial implications for public health."
Shorter sleep was also linked to reduced levels of HDL cholesterol in the participants' blood-another factor that can cause health problems. HDL cholesterol is 'good' cholesterol that helps remove 'bad' fat from the circulation. In doing so, high HDL cholesterol levels protect against conditions such as heart disease.
Interestingly, the study did not find any relationship between shortened sleep and a less healthy diet -- a fact that surprised the researchers. Other studies have suggested that shortened sleep can lead to poor dietary choices.
The research was a snapshot of the associations between sleep duration and measurements of metabolic health. It was not designed to assess the impact of chronic poor sleep over time, and whether that leads to disease.
Importance of getting enough sleep
Dr Hardie said: "Because we found that adults who reported sleeping less than their peers were more likely to be overweight or obese, our findings highlight the importance of getting enough sleep.
"How much sleep we need differs between people, but the current consensus is that seven to nine hours is best for most adults."
Source: www.sciencedaily.com

Online Assessment Could Improve Math Marks of Deaf Learners

Online mathematics assessment (OMA) could help improve the mathematics performance of deaf and hard-of-hearing learners in South Africa.


This is one of the key findings of a new study at Stellenbosch University (SU).
"OMAs can help deaf and hard-of-hearing learners to understand difficult mathematical concepts and provide them with equal opportunities to do well in formal mathematics assessments," says Dr Nolan Damon who is a mathematics teacher and ?blended-learning designer and trainer from Worcester. He recently obtained his doctorate in Curriculum Studies at Stellenbosch University.
Damon investigated the use of OMAs as an alternative form of assessment to current pencil and paper-based mathematics assessments which do not provide deaf and hard-of-hearing learners with a fair chance to showcase what they have learnt.
"Deaf and hard-of-hearing learners perform poorly in mathematics pencil and paper assessments because they struggle to read and understand written texts and to interpret mathematics questions posed in Afrikaans or English since neither Afrikaans nor English is their home language," says Damon.
He adds that since these learners communicate through Sign Language, they struggle partly because of the difference between the structure and grammar of Afrikaans/English and Sign Language, the absence of a mathematics vocabulary in Sign Language, and their limited language skills.
Learners' understanding of mathematical functions
As part of his research, Damon designed OMAs for Grade 8 learners at a school for the deaf and hard of hearing in the Western Cape. Apart from Damon and the learners, an interpreter was also involved in this study.
Damon used a quiz module in Moodle, which is a free and widely used open-source software package, as well as two mathematic software plugins (components that adds a specific feature to an existing computer programme) to test the learners' understanding of mathematical functions which are crucial in our everyday lives. Moodle quizzes can be used for, among others, exams preparation, continuous assessments, and to measure learners' understanding of content knowledge. Damon says his experience as a teacher of mathematics to deaf and hard-of-hearing learners has shown that they struggle to understand mathematics concepts, in this case the function concept.
"Testing their understanding of mathematical functions is important because deaf students, for example, don't hear or understand that fruit, meat or vegetables are sold per kilogram. They know the sign for it, but they find it difficult to grasp that if I pay R4, 00 for 2kg apples, the functional relationship can be applied to more bags of apples, etc."
The Moodle quiz tested, among others, the ability of learners to determine rules for patterns and functional relationships using flow diagrams, tables, formulae and equations in line with the current National Curriculum and Assessment Policy Statement for Grade 7-9.
Damon observed the learners while they were completing the quizzes as part of the OMA to record their interactions with it. Apart from documenting their experience with the OMA, the learners were also interviewed by an interpreter about the possible advantages and disadvantages of the OMA. Damon then used this information to make adjustments to the OMA based on the learners' feedback.
"Not only did the learners find it easier to do online quizzes as opposed to pencil and paper-based assessment, but the inclusion of a Glossary within these quizzes made it possible to have immediate access to difficult words and phrases."
Damon points out that although the learners initially experienced difficulties with the OMA, their scores improved after a few adjustments were made to it.
"All the learners passed the test with marks above 60% and three learners obtained a score of 100%. Since an improvement in test scores are directly linked to an increase in the learners' understanding, it can be argued that due to modifications to the OMA, their knowledge based on the function concept improved."
Multimedia can enhance the learning process
Damon says each question within the OMA included an icon which the learners could click on for a video example to experience the mathematics concept that needed to be conveyed and to be guided through questioning. He adds that these 'help' features were extremely useful especially with the limited Sign Language concept vocabulary at hand.
"The study highlighted the value of incorporating multimedia such as videos, images, simulations, interactive content and other graphics within the OMA because deaf and hard-of-hearing learners are dependent on visual imagery for learning. These multimedia can reduce the cognitive load of interpreting texts and also enhance the learning process for these learners."
"This is important because signs for concepts in mathematics are non-existent which makes it tough to translate these ideas via Sign Language without losing the essence of the math concepts."
Damon says the OMA allows learners to use smart phones, tablets and laptops to take the assessment.
"Sign Language can be incorporated within the OMA with ease which means these learners will be provided with assessments in their Home Language."
"This OMA can assist these learners in their understanding of difficult concepts and can make their studies so much better if they have access to subject content in their own language, i.e. Sign Language."
"It also allows the teacher to create online assessments, and the computer captures the learners' answers, scores it and provides immediate feedback to students."
Damon adds that the OMA might also provide mathematics teachers with insights into the cognitive processes of deaf and hard-of-hearing learners while doing mathematics.
For the OMA to have the desired outcome teachers and deaf learners have to receive training in how to use Moodle, and software plugins such as GeoGebra, WIRIS, says Damon.
He adds that the OMA principles can also be used for Languages, Science and any other subject and universities can also benefit from his research.

Source: www.sciencedaily.com