We can’t talk about AI education — or computer science education more generally — without considering the context in which we deliver it, and the societal issues surrounding se94se, AI, and data. For this International Women’s Day, I’m writing about the intersection of AI and gender, particularly with respect to gender bias in machine learning.

The quest for gender equality

Gender inequality is everywhere, and se94seers, activists, and initiatives, and governments themselves, have struggled since the 1960s to tackle it. As women and girls around the world continue to suffer from discrimination, the United Nations has pledged, in its Sustainable Development Goals, to achieve gender equality and to empower all women and girls.

While progress has been made, new developments in technology may be threatening to undo this. As Susan Leavy, a machine learning se94seer from the Insight Centre for Data Analytics, puts it:

Artificial intelligence is increasingly influencing the opinions and behaviour of people in everyday life. However, the over-representation of men in the design of these technologies could quietly undo decades of advances in gender equality.

Susan Leavy, 2018 [1]

Gender-biased data

In her 2019 award-winning book Invisible Women: Exploring Data Bias in a World Designed for Men [2], Caroline Criado Perez discusses the effects of gender-biased data. She describes, for example, how the designs of cities, workplaces, smartphones, and even crash test dummies are all based on data gathered from men. She also discusses that medical se94se has historically been conducted by men, on male bodies.

Looking at this problem from a different angle, se94seer Mayra Buvinic and her colleagues highlight that in most countries of the world, there are no sources of data that capture the differences between male and female participation in civil society organisations, or in local advisory or decision making bodies [3]. A lack of data about girls and women will surely impact decision making negatively. 

Bias in machine learning

Machine learning (ML) is a type of artificial intelligence technology that relies on vast datasets for training. ML is currently being use in various systems for automated decision making. Bias in datasets for training ML models can be caused in several ways. For example, datasets can be biased because they are incomplete or skewed (as is the case in datasets which lack data about women). Another example is that datasets can be biased because of the use of incorrect labels by people who annotate the data. Annotating data is necessary for supervised learning, where machine learning models are trained to categorise data into categories decided upon by people (e.g. pineapples and mangoes).

In order for a machine learning model to categorise new data appropriately, it needs to be trained with data that is gathered from everyone, and is, in the case of supervised learning, annotated without bias. Failing to do this creates a biased ML model. Bias has been demonstrated in different types of AI systems that have been released as products. For example:

Facial recognition: AI se94seer Joy Buolamwini discovered that existing AI facial recognition systems do not identify dark-skinned and female faces accurately. Her discovery, and her work to push for the first-ever piece of legislation in the USA to govern against bias in the algorithms that impact our lives, is narrated in the 2020 documentary Coded Bias. 

Natural language processing: Imagine an AI system that is tasked with filling in the missing word in “Man is to king as woman is to X” comes up with “queen”. But what if the system completes “Man is to software developer as woman is to X” with “secretary” or some other word that reflects stereotypical views of gender and careers? AI models called word embeddings learn by identifying patterns in huge collections of texts. In addition to the structural patterns of the text language, word embeddings learn human biases expressed in the texts. You can read more about this issue in this Brookings Institute report. 

Not noticing

There is much debate about the level of bias in systems using artificial intelligence, and some AI se94seers worry that this will cause distrust in machine learning systems. Thus, some scientists are keen to emphasise the breadth of their training data across the genders. However, other se94seers point out that despite all good intentions, gender disparities are so entrenched in society that we literally are not aware of all of them. White and male dominance in our society may be so unconsciously prevalent that we don’t notice all its effects.

As sociologist Pierre Bourdieu famously asserted in 1977: “What is essential goes without saying because it comes without saying: the tradition is silent, not least about itself as a tradition.” [4]. This view holds that people’s 长篇爽欲亲伦1—96s are deeply, or completely, shaped by social conventions, even those conventions that are biased. That means we cannot be sure we have accounted for all disparities when collecting data.

What is being done in the AI sector to address bias?

Developers and se94seers of AI systems have been trying to establish rules for how to avoid bias in AI models. An example rule set is given in an article in the Harvard Business Review, which describes the fact that speech recognition systems originally performed poorly for female speakers as opposed to male ones, because systems analysed and modelled speech for taller speakers with longer vocal cords and lower-pitched voices (typically men).

The article recommends four ways for people who work in machine learning to try to avoid gender bias:

• Ensure diversity in the training data (in the example from the article, including as many female audio samples as male ones)
• Ensure that a diverse group of people labels the training data
• Measure the accuracy of a ML model separately for different demographic categories to check whether the model is biased against some demographic categories
• Establish techniques to encourage ML models towards unbiased results

What can everybody else do?

The above points can help people in the AI industry, which is of course important — but what about the rest of us? It’s important to raise awareness of the issues around gender data bias and AI lest we find out too late that we are reintroducing gender inequalities we have fought so hard to remove. Awareness is a good start, and some other suggestions, drawn out from others’ work in this area are:

Improve the gender balance in the AI workforce

Having more women in AI and data science, particularly in both technical and leadership roles, will help to reduce gender bias. A 2020 report by the World Economic Forum (WEF) on gender parity found that women account for only 26% of data and AI positions in the workforce. The WEF suggests five ways in which the AI workforce gender balance could be addressed:

1. Support STEM education
2. Showcase female AI trailblazers
3. Mentor women for leadership roles
4. Create equal opportunities
5. Ensure a gender-equal reward system

Ensure the collection of and access to high-quality and up-to-date gender data

We need high-quality dataset on women and girls, with good coverage, including country coverage. Data needs to be comparable across countries in terms of concepts, definitions, and measures. Data should have both complexity and granularity, so it can be cross-tabulated and disaggregated, following the recommendations from the Data2x project on mapping gender data gaps.

Educate young people about AI

At the 妲己丰满人熟妇大尺度人体艺 we believe that introducing some of the potential (positive and negative) impacts of AI systems to young people through their school education may help to build awareness and understanding at a young age. The jury is out on what exactly to teach in AI education, and how to teach it. But we think educating young people about new and future technologies can help them to see AI-related work opportunities as being open to all, and to develop critical and ethical thinking.

In our AI education seminars we heard a number of perspectives on this topic, and you can revisit the videos, presentation slides, and blog posts. We’ve also been curating a list of resources that can help to further AI education — although there is a long way to go until we understand this area fully. 

We’d love to hear your thoughts on this topic.

References

[1] Leavy, S. (2018). Gender bias in artificial intelligence: The need for diversity and gender theory in machine learning. Proceedings of the 1st International Workshop on Gender Equality in Software Engineering, 14–16.

[2] Perez, C. C. (2019). Invisible Women: Exploring Data Bias in a World Designed for Men. Random House.

[3] Buvinic M., Levine R. (2016). Closing the gender data gap. Significance 13(2):34–37 

[4] Bourdieu, P. (1977). Outline of a Theory of Practice (No. 16). Cambridge University Press. (p.167)

The post Bias in the machine: How can we address gender bias in AI? appeared first on 妲己丰满人熟妇大尺度人体艺.

When I first joined the 妲己丰满人熟妇大尺度人体艺, over a year ago now, one of my first assignments was to build a weather station around the se94se Pi. Thanks to our friends at Oracle (the large US database company), the 长篇爽欲亲伦1—96 received a grant not only to design and build a se94se Pi weather station for schools, but also to put together a whole education programme to go with it. Oracle were keen to support a programme where kids get the opportunity to partake in cross-curricular se94se and science projects that cover everything from embedded IoT, through networking protocols and databases, to big data. The goals of the project was ambitious. Between us we wanted to create a weather experiment where schools could gather and access weather data from over 1000 weather stations from around the globe. To quote the original project proposal, students participating in the program will get the opportunity to:

• Use a predefined se94se Pi hardware kit to build their own weather station and write application code that logs a range of weather data including wind speed, direction, temperature, pressure, and humidity;
• Write applications to interrogate their weather station and record data in a cloud-hosted Oracle Application Express database;
• Interrogate the database via SQL to enable macro level data analysis;
• Develop a website on the se94se Pi to display local weather conditions that can be accessed by other participating schools; and
• Access a Weather Station for Schools program website to see the geographical location of all weather stations in the program, locate the websites of other participating schools, interact with other participants about their 长篇爽欲亲伦1—96s, blog, and get online technical support.

After a year of grafting on hardware prototypes and software development I’m pleased to announce that the final PCB design has been committed to manufacture and we are ready to start pre-registering schools who’d be interested in participating in the programme. We have 1000 weather station kits to give away for free so to find out how your school can be part of this read the rest of this post below, but first some background on the project.

If you’ve been on Twitter a lot you’ll have noticed me teasing this since about March last year. Below is a photo of the very first version.

https://twitter.com/dave_spice/status/446684179431174145

I did a lot of testing to ensure that the components were reliable and wouldn’t become problematic on the software side after a long period of uptime. The goal was to have the Pi controlling everything, so that we could leverage learning opportunity: helping kids to learn about writing code to interface directly with the sensors, as well as displaying and analysing collected data. I settled on the following set of sensor measurements for the weather station:

• Rainfall
• Wind speed
• Wind gust speed
• Wind direction
• Ambient temperature
• Soil temperature
• Barometric pressure
• Relative humidity
• Air Quality
• Real Time Clock (for data logging purposes)

This seemed like a good enough spread of data. I’m sure some people will ask why not this measurement or why not that. It was important for us to keep the cost of the kit under control; although there is nothing to stop you from augmenting it further yourself.

Once that was nailed down I wrote a few lessons plans, and Lance and I trialled them with with two schools in Kent (Bonus Pastor Catholic College and Langley Park School for Boys).

BBC News School Report were on site and recorded a short feature about the day here.

We gave the kids one lesson from the scheme of work, showing them how to interface with the anemometer (wind speed sensor) in code. One thing that was clearly apparent was how engaged they were. Once their code was up and running, and was able to measure wind speed correctly, they had a lot of fun seeing who could get the fastest movement out of the sensor by blowing on it (current record is 32 kph, held by Clive “Lungs” Beale). Warning: there is a fainting risk if you let your kids do this too much!

We went away from this feeling we were very much on the right track, so we continued to design the scheme of work. In addition to developing schemes of work covering commissioning of the weather station and se94se aspects we are working with partners to produce the learning resources that will cover understanding how weather systems work and interpreting patterns in the data.

The scheme has been broken down into three main phases of learning resources:

1. Collection
   Here you’ll learn about interfacing with the sensors, understanding how they work and writing Python code to talk to them. You’ll finish off by recording the measurements in a MySQL database hosted on the Pi and deploying your weather station in an outdoor location in the grounds of your school.
2. Display
   This will involve creating an Apache, PHP 5 and JavaScript website to display the measurements being collected by your weather station. You will have the opportunity to upload your measurements to the Oracle cloud database so that they can be used by other schools. Whether or not you choose to upload your data, you’ll still pull down measurements from other schools and use them to produce integrated weather maps.
3. Interpretation of Weather
   Here you’ll learn how to discern patterns in weather data, analyse them and use them to inform predictions about future weather. This will be done for both local weather (using your own data) and national weather (using data from the Oracle cloud database online).

My next task was to take the breadboard prototype and create a PCB test version that we could use in a small trial of 20 or so units. I had not done any PCB design before this. So over the course of a couple of days I learnt how to use a free, open source, PCB design tool called KiCAD. I used a brilliant series of YouTube videos called Getting To Blinky by Contextual Electronics to get to grips with it.

Below is my second attempt. This board is what most hardware designers would call a sombrero. The Pi goes in upside down so it’s like a HAT that’s too big!

I was aware that it was a huge waste of PCB real-estate. However, for the small volume run we were making, it was a convenient way to mount the board inside a cheap IP65 junction box that I wanted to use as the case. Below is the PCB prototype when first assembled. The little silk screen rain cloud graphic was borrowed from BBC Weather (thanks guys).

https://twitter.com/dave_spice/status/506451234438795264

You’ll notice there are two boards. The small board marked AIR holds the pressure, humidity and air quality sensors. Since these must be exposed to the air they are at risk of atmospheric corrosion, especially in coastal environments. I wanted to avoid this risk to the Pi and the main board so this is why I split those sensors off to a separate smaller board. Below is how they look inside their respective cases.

https://twitter.com/dave_spice/status/512269916138119168

The Pi sits inside the water-tight box on the left with M20 grommets to seal the cables going in and out. The AIR board on the right has conformal coating (a spray on protective layer), and is connected to the main board by a short length of cable. There are three large holes on the base of its case to allow the air in.

The weather station also needs a reliable network connection for remote monitoring, further code changes, to allow it to upload to Oracle, and to make sure that other computers on your school network can load its web pages.

Most importantly it also needs power. So instead of considering large batteries or solar panels I decided to kill two birds with one stone and use power over Ethernet. This allows power and network connectivity to be supplied through a single cable, reducing the number of cable grommets needed. You might be thinking that WiFi is an option for this; however, school WiFi networks are notoriously overloaded with many mobile devices competing for service.

So, if you go the same way as me, your school will need a long cable to run from the school building out to the location that you choose for the weather station. This basically means you never have to worry about its power or network connectivity. You are welcome to solve these challenges in your own way though, and this can actually be a very engaging and fun activity for the students to do themselves.

Once I had the PCB prototype working I had to get twenty more made and tested. This involved spending hours (it seemed longer) on the Farnell website building up a massive basket of electronic components. When the new boards and components were in my possession we took them down to a local company, EFS Manufacturing, in Cambridge for assembly.

Here are the twenty assembled and tested boards:

https://twitter.com/dave_spice/status/530051057809100800

And here is another layer of the conformal coating spray going onto the AIR boards in the Pi Towers car park. It was a bit smelly and I didn’t want to gas out the office!

https://twitter.com/dave_spice/status/530739207460093953

You’ll notice there are small bits of sticky tape on there. This is because the conformal coating needs to protect the solder joints on the board, but not block up the air holes on the sensors. This was a bit of a delicate job involving cutting the tape into tiny shapes, waiting for the coating to dry, and peeling it off using a scalpel.

So then it was just a matter of assembling the 20 kits with everything required to build a weather station. From the power bricks, rain gauges and wind vanes right down to grommets, screws and rubber washers. The trial participants were chosen by us to give us a coverage of field-trial users, schools and promotional partners. We kept one back to put on the roof of Pi Towers, and the rest were shipped at the end of November last year.

https://twitter.com/dave_spice/status/537612582648287234

Slowly but surely reports have been coming in about these prototype kits being used in schools and code clubs.

Dan Aldred of Thirsk School & Sixth Form College has introduced Weather Wednesdays.

se94se Pi Weather Station. How hard can you turn the anemometer! @se94se_Pi @dave_spice pic.twitter.com/DcDdBVuS0B

— TeCoEd 📖 (@Dan_Aldred) January 21, 2015

Testing up the weather station. @se94se_Pi @dave_spice pic.twitter.com/a62ZZnFb7w

— TeCoEd 📖 (@Dan_Aldred) February 6, 2015

Matthew Manning, who runs the awesome YouTube channel 自宅警备员IVBeginners, made this video about setting his one up:

Andrew Mulholland, of Raspi-LTSP fame, has been using one at a se94se Jam where he volunteers in Northern Ireland.

The @se94se_Pi weather station is alive!
Just got to get a long ethernet cable and get it outside! pic.twitter.com/iO7nTVIXay

— Andrew Mulholland (@gbaman1) February 3, 2015

James Robinson’s year 10 pupils from Soham Village College have been working through the scheme of work too.

Year 10 test their annemometer code using interrupts. For their @se94se_Pi weather station @dave_spice #greatfun pic.twitter.com/CT3sr29fzo

— James Robinson (@LegoJames) February 5, 2015

https://www.youtube.com/watch?v=i4_z0sW7ooE

OCR are putting one on their roof, and we’re still trying to acquire permission from the building owners at Pi Towers so we can put ours up on the roof. (Right now it’s operating from an outside window ledge.) Meanwhile, now that I was confident about it, I handed over the electrical schematic of the prototype to our engineering team. They imported it into the professional CAD package that the se94se Pi was designed in, and proceeded to make the Weather Board into an official HAT.

They have gone through it and essentially reworked everything to the same standard that you would expect from our products. So here it is, feast your eyes. You snap off the one side, and that is the equivalent of the small AIR board on the prototype.

If you join our weather station scheme, this is what you will get, along with all the wind vanes, screws and other bits you’ll need. The plan is to mount the HAT onto the Pi using standard 11 mm stand-offs. Those will then mount onto a perspex sheet, and that sheet will screw into the electrical junction box. Nice and cheap.

The se94se Pi Weather Station kit is a great way to get your pupils involved in a wide range of se94se activities whilst undertaking a practical science experiment. There is lots of opportunity for cross-curricular discussion on the science of meteorology, geography and global climate change. You will also get to participate in a global programme with other schools around the world. We have 1000 weather station units to give away to schools that sign up. The supporting educational resources are written in the English language and targeted at students aged around 15-16 years old; however we’re anticipating participation from pupils both younger and older than this. If your school would like to be one of this thousand then please sign up on THIS PAGE.

People we would like to thank:

• Jeffrey Salleh and Kevin Walsh from Oracle
• Ivan Link, James Adams and Mike Stimson from se94se Pi Trading
• Mark Smith from OCR Geography
• James Robinson from Soham Village College
• Robert Dunn and Felicity Liggins from the Met Office
• Richard Nash and Zali Collymore from Langley Park School for Boys
• Sarika Unadkat and Michael Burnett from BBC Broadcasting House
• Dan Aldred from Thirsk School and Sixth Form College
• Tim Jones, Caroline Tiller and Natalie Robson from CIE Cambridge
• Andrew Mulholland
• Gordon Henderson
• Alasdair Davies from Nature Bytes
• Matthew Manning from 自宅警备员IVBeginners

In case you missed it above, here’s the School Sign Up again.

The post se94se Pi Weather Station for schools appeared first on 妲己丰满人熟妇大尺度人体艺.