News Channels: LEARN

During the five years that Paula Cacault commuted to Argentina’s largest public university, where she studied economics, she made sure to take a folding stool in her backpack in case she couldn’t find a seat in the crowded lecture hall. This is one of the reasons why the use of distance learning to alleviate the problem of overcrowded university classrooms is a matter of concern for the head of operations at the EPFL’s Enterprise for Society Center (E4S).

In October 2023, the study that she carried out with Christian Hildebrand, a professor of marketing analytics at the University of St. Gallen, and Jérémy Laurent-Lucchetti and Michele Pellizzari, both economics professors at the University of Geneva, won the Swiss Education Research Prize. In this research, they looked specifically at live streaming and how it can affect both in-person attendance and student achievement.

The research was performed in 2017 and involved 1,459 Bachelor’s students in the University of Geneva’s economics and management program. They were randomly assigned to three groups: 15% of the students were never given access to the live streaming service for their classes, 15% of the students always had access and 70% of the students could access the service during only some weeks selected at random. “We thought students would rush to use the streaming service whenever they could,” says Cacault. “But as it turned out, just 10% of them took advantage of it.” The authors subsequently analyzed students’ exam grades, and found that the live streaming generally improved the grades of high-achieving students and lowered the grades of low-achieving ones. “The literature indicates that students who are already struggling tend to do worse when professors stray from conventional teaching methods,” says Pellizzari.

Evolving habits

Fast-forward six years, distance learning has become much more widespread and students’ habits have changed. To get a better grasp of how EPFL students prefer to learn, Pierre Dillenbourg, the School’s Associate Vice President for Education, ran three surveys in conjunction with EPFL’s Teaching Support Center (CAPE) and LEARN Center. “Professors had noticed a shift in the way students approach their classes,” he says. “We wanted to understand what was going on so that we could adapt our teaching methods in response.”

The three surveys were carried out in June 2022 (2,700 respondents), November 2022 (2,900 respondents) and June 2023 (1,400 respondents). The first two surveys found that of the students who were already at EPFL before the pandemic, over half (62.6% in June and 51.6% in November) said they now attended class in person less often.

Another finding was that in-person attendance proved to be higher in the fall. In November 2022, 74.7% of respondents said they “always” or “often” attended a class in person, compared with 57.4% in June 2022 and 61.7% in June 2023.

Even when students viewed lectures remotely, over half of them (52% on average for the three surveys) said they “always” or “often” came to campus anyway – mainly to study, work on projects or take part in social events.

Recorded lectures are a tool just like any other

While students appreciate it when professors make recordings of their lectures available online, that doesn’t prevent them from attending the class in person. “We were curious to see how students kept on top of the subject matter when they didn’t come to class, and we found that they used handouts, textbooks and slides just as much as they used recorded lectures,” says Dillenbourg. “So these recordings aren’t a decisive factor for in-person attendance.”

Students typically go to classes they find interesting, or particularly difficult, or stimulating because the professors run activities that aren’t part of the recorded lectures. The pandemic appears to have raised the bar in terms of what students expect from their teachers. Today, they want interactive experiences where there’s real added value to being physically present in a classroom. But has this selective attitude towards attendance impacted skills acquisition?

We were curious to see how students kept on top of the subject matter when they didn’t come to class, and we found that they used handouts, textbooks and slides just as much as they used recorded lectures. So these recordings aren’t a decisive factor for in-person attendance.

“The surveys were anonymous, so we couldn’t establish a link between reported in-person attendance and exam grades,” says Dillenbourg. “However, studies in other countries have shown that the correlation is very weak, except for first-year students and for classes that involve hands-on experiments. But of course, our goal at EPFL is for students to take full advantage of our great campus to learn from each other and interact with their professors.”

In terms of whether these new learning habits have widened the gap between high-achieving and low-achieving students, Cacault believes another study is in order. “Our 2017 study found that when there is a high cost associated with getting to class – a long commute or bad weather, for example – both categories of students use live streaming to roughly the same extent,” she says. “Why live streaming improves the grades of those who are doing well and lowers the grades of those who are struggling? We imagine it relates to the fact that when live streaming isn’t available, struggling students will make the effort to come to class even when the cost is high, while the best students will be tempted to skip class and learn on their own. When live streaming is available, however, struggling students will watch the lecture online instead of going to class, which hurts their grades, but the best students will watch the lecture instead of learning on their own, and this lifts their grades.”

The 2017 study only looked at the effects of live streaming – which isn’t the same as making recorded lectures available online. To examine this latter practice, Laurent-Lucchetti and Pellizzari are planning another study at the University of Geneva. “We’re still in the early stages, but our objective is to determine whether having access to video recordings affects in-person attendance and student achievement,” they say.

Current hot topics, such as budget cuts and the preservation of the quality of education in light of the influx of new students at the Bachelor’s level, dominated the majority of presentations delivered by the President of the ETH Board, Professor Michael Hengartner, and by the rector of ETH Zurich, Professor Günther Dissertori.

A success that is not sustainable

In late January, the Federal Council announced new cost-cutting measures, reducing the contribution allotted to the EPFs for the 2025 budget by 100 million francs. A decision which undoubtedly puts EPFL and ETH Zurich under pressure.

“Politicians often ask me what the secret of our success is,” Professor Michael Hengartner shared during EPFL’ General Assembly of the Teacher Council (CCE). According to the President of the strategic management and supervisory body of the ETH Domain, the success of both institutions relies on three pillars: openness to the world, autonomous governance, and sound finances. “We want to create conditions in which the best students and scientists feel compelled to choose us,” he declared, “and today we have challenges to address on several of these pillars.”

Professor Hengartner also emphasized the imperative of continuing efforts to increase diversity and inclusion, especially by training more women, and placing education at the forefront of current considerations, stressing the need to preserve the quality of education above all.

“The questions we need to ask ourselves are: ‘What do we want to teach the next generation? What skills, what knowledge do we want to impart? And finally, to whom do we want to teach? The number of women in our schools is increasing very slowly. If we trained as many women as men in computer science, for example, Switzerland would not lack female computer scientists.”

Similar observations

From the perspective of ETH Zurich, Professor Günther Dissertori draws a similar conclusion: both institutions are victims of their success. For the rector of EPFL’s “big sister,” the key is to listen to stakeholders and deliver training in line with their evolving needs. He also pointed at the importance of developing project-based learning to promote a more holistic and systemic approach and to enhance students’ ability to formulate approaches rather than finding predefined answers. Finally, he suggested considering new modalities for continuing education as another potential avenue. “One way to understand the concrete needs of the industry is project-based teaching. There is a growing demand for shorter courses, micro-credit systems. We must increasingly listen to our former students, industry, and society.”

Addressing the gender imbalance in the STEM fields, Günther Dissertori encouraged an evidence-based approach and called for intensified analysis of the impact of the many existing programs aimed at remedying this problem. In particular, he cited both schools’ programs in the learning sciences as a potential solution.

“Everything isn’t going well, but we have good news”

These talks paved the way for a roundtable discussion on the “Challenges and Opportunities for Engineering Education in Switzerland” during the second part of the morning. Moderator and journalist Alexis Favre invited representatives from EPFL to join Michael Hengartner and Günther Dissertori on stage, calling on the Associate Vice President for Education, Pierre Dillenbourg, the Vice President for Responsible Transformation, Professor Gisou van der Goot, the Director of the Machine Learning for Education Laboratory (ML4ED), Professor Tanja Käser, and last but not least, representing the student body, Jakub Frybes, co-president of AGEPoly.

Opening the discussion, Pierre Dillenbourg called for not succumbing to pessimism. “The spirit of our meetings is gloomier than reality,” he tempered. According to him, alongside emerging challenges, encouraging advances also deserve recognition. These include the recent appointment of the first two “MER Type II”, a title that allows recognizing individuals who have taught for years, innovated, and developed resources “but who are not on a full professorship track.” Adding to this, Dillenbourg also mentioned EPFL’s transition to a new academic management system, the infrastructure development project that will eventually offer 1,500 new learning spaces adapted to project-based teaching, and a budget of 500,000 francs allocated to Student Assistants. “Everything isn’t going well, but we do have good news,” he summed up.

Collaboration and communication key to addressing current challenges

In a tumultuous context then, experts from both Federal Institutes of Technology emphasized the importance of adapting to the evolving landscape while maintaining their core values and commitment to the quality of education. The focus was on the need to continue developing students’ transversal skills and to invest in close collaboration between the two federal institutions.

“We have two joint master’s programs in cybersecurity and nuclear engineering, a joint center for data science, we are discussing and finalizing a joint center for AI, and we are discussing another center in the near future dedicated to the learning sciences. So, we are trying to have more and more collaboration; we have recently discussed having joint hybrid lessons so that we can benefit from courses that exist in one school but not in the other,” explained Pierre Dillenbourg.

For Professor Gisou van der Goot, head of the new Vice Presidency for Responsible Transformation, ongoing societal changes should be seen as an opportunity to challenge the status quo, rethink curricula, and not lose sight of the values at the heart of universities. “Universities are places where we should be able to debate everything. We must defend our intellectual and moral autonomy,” she emphasized.

The remainder of the Teaching Day gave the floor to EPFL teachers through a series of discussion groups to explore the issues at hand. Participants also exchanged views on more targeted themes such as AI tools for education, or broader societal topics such as mental health, cross-cutting skills, sustainability, diversity, equity, and inclusion.

While the necessity of understanding AI’s inner workings and grasping the full scope of what is at stake with these technologies seems to have primarily concerned the so-called “active” population, retirement does not necessarily imply complete disconnection from such topics.

In Switzerland, senior citizens are far from lagging behind and are, on the contrary, quite tech-savvy as it turns out. A recent study conducted by Pro Senectute suggests just that, indicating a 95 percent internet usage among 65 to 69-year-olds.

Additionally, as with most technologies, AI applications extend well beyond the professional realm, inevitably weaving their way into our private lives. Therefore, the growing prevalence of AI indicates that it should be a topic of interest for society as a whole. However, to join the conversation one must first understand its codes. To bridge this generational gap, an initiative emerged through collaboration between the senior club “Toujours Plus” of Migros Vaud and our team of experts at the Center LEARN. As a result, two three-hour workshops on AI were organized for its members this autumn.

Tailored training

To address this new challenge, the training team comprised of Sonia Agrebi, Project Manager, Felipe Martinez, Project Manager, and Frédérique Chessel Lazzarotto, Coordinator of external missions at the Center LEARN, opted for an original and tailored approach. “We were not familiar with the skill levels or expectations of this audience that is new to us, so we chose accessible themes and an engaging approach to ease them into discovering the basic concepts of artificial intelligence,” explained Frédérique Chessel Lazzarotto.

The trainers thus guided the 44 participants, dispelling misconceptions and demystifying common AI tools one by one before delving into the history of machine learning, going as far back as the 1950s. Then, using an imaginative mushroom-picking scenario, the training team skillfully shed light on the main paradigms of ML. The workshop concluded with practical exercises where everyone had the opportunity to experiment with tools such as ChatGPT or Dall-E to generate recipes and greeting cards.

A gratifying outcome

The Center LEARN analyzed this training to evaluate its effectiveness. The post-training feedback was extremely positive: 89% of participants expressed great satisfaction with the content covered, while 78% praised the interactive teaching methods adapted to their pace.

The most striking aspect was the observed change in attitude towards AI among the participants. Before the training, they harbored some mistrust towards these tools, but after the sessions, wonder largely outweighed fear. “Their desire to understand a sometimes complex topic was manifested through pertinent questions, rooted in thoughtful personal reflections. The attentive listening and remarkable commitment of each participant contributed to enriching this training experience,” said Felipe Martinez.

These workshops also altered their intentions regarding the use of AI. While 84% of participants declared that they were not using this technology at the start of the day, now 57% are considering it, reflecting the positive impact of the workshops on their perspectives and willingness to explore new horizons.

Before the workshop, 80% admitted not knowing about AI, a percentage that dropped to 2% after the course. Concurrently, the percentage of participants claiming to understand AI soared from 20% to 98%, demonstrating the effectiveness of the training in deepening their knowledge of the subject. “Defining artificial intelligence and exploring its diverse application domains not only demystified what AI is but also revealed that it was already very much present in their daily lives, often in completely unsuspected ways,” explained Sonia Agrebi.

These results highlight the effectiveness and relevance of the pedagogical approach tailored to seniors proposed by our team and pave the way for the implementation of other similar interventions in the future.

When Tanja Käser landed the role of tenure track assistant professor at EPFL, she never imagined she’d spend her first day in the job alone, working from a makeshift desk in the corner of her bedroom. That was in May 2020, at the height of the pandemic. “It was strange in this situation to tell myself that I was now a professor at EPFL,” she smiles. “I didn’t meet my new colleagues face-to-face for several months.” That fall, she taught her first classes online. And as spring rolled in, her daily routine remained unchanged: giving lectures to a screen teeming with faces and black rectangles bearing still-unfamiliar names.

For the head of the EPFL’s Machine Learning for Education Laboratory (ML4ED), it was an unconventional start to life as an educator. But it helped her develop a solid teaching ethos from the outset – one based on active learning, repetition, and linking theory to practice. “In my view, to be a good teacher, you have to be able to ‘read’ a class and adapt your approach to each student’s strengths and learning style,” she says. This guiding philosophy earned her the Credit Suisse Award.

A purpose-driven approach

“I’m fascinated by the way humans learn,” says Tanja Käser, whose research involves developing machine-learning models to understand and improve human learning. “For me, it’s really important to do something meaningful. I also like the ethical side of things, which is especially important when you’re applying algorithms to education.”

In my view, to be a good teacher, you have to be able to ‘read’ a class and adapt your approach to each student’s strengths and learning style.

Despite her interest in human learning, she never saw herself pursuing a career in teaching. “Quite the opposite in fact!” she says with a smile. “I come from a family of teachers, so I was dead set against the idea.” However, she always knew she wanted to do something purposeful and make a difference through her work. So it was only natural that she took her interest in computer science – something she came to by way of civil engineering – and applied it to the field of education. “During a week-long taster program for girls at ETH Zurich, I saw these amazing virtual demonstrations of civil engineering projects,” says Käser. “They really piqued my curiosity.” After completing her studies, she spent a year working as a consultant at McKinsey before returning to research.

Holding the attention of 600 students

As a researcher passionate by education, Tanja Käser is also intensely committed to her courses. Especially the weekly discrete mathematics class she’s teaching for the second time this fall, which covers some of the key concepts behind algorithms. She acknowledges that lecturing to 600 first-year students packed into a Rolex Forum auditorium isn’t ideal, and that coordinating a team of 30 student assistants and 10 PhD assistants is an energy-sapping task. So how does Käser hold the attention of hundreds of people when it’s impossible to catch everyone’s eye? “I try to speak for no more than 10 or 15 minutes at a time,” she says. “Then I ask questions using the SpeakUp app. Students discuss the questions among themselves, which tells me whether they’ve understood the material. Each semester, I also give four quizzes to make sure my students are on track. It seems to work well. I’m very happy because setting up this class took a lot of effort.”

She admits that work consumes a lot of her time, including at weekends. But that’s because she finds endless fascination in exploring algorithms and delving into the nuts and bolts of human learning. For her Master’s class – Machine Learning for Behavioral Data – Tanja Käser uses a project-based approach implemented in association with the Swiss EdTech Collider, a community of educational technology startups.

“For each Master’s class, we select two or three startups that are willing to share their data with us,” she explains. “Then we divide the students into groups of three, and they choose the dataset they want to work with. The startups benefit from having the students analyze their data, while the students get a chance to apply what they’ve learned by tackling concrete problems and doing something useful.” At the end of the semester, each group give a poster presentation to the startups and the teaching team. A stimulating moment of exchange. Fortunately, in stark contrast to her early days alone in front of her screen.

Traduit du français

On May 11^th 2023, a kick-off event gathered the program’s key stakeholders, as well as all EPFL scholars for the launch of this enhanced learning program that is based on four main dimensions: research, leadership development, social engagement and networking. Selected for their motivation, leadership potential and research curiosity, these scholars also have demonstrated strong convictions that change must happen to make the world a better place. “During the recruitment interviews, I was impressed by the values and degree of maturity some of the students already exhibit” says Dr. Sabrina Rami-Shojaei, Director of EPFL’s Education Outreach Department and Member of the Laidlaw Admission Committee.

I hope that this program will successfully support them in their efforts and push them to make decisions guided by ethics, integrity and empathy.

PFL has joined forces with the Laidlaw Foundation to offer its Bachelor’s students a unique opportunity to complement their academic journey with a comprehensive curriculum that includes personal development training focusing on transversal skills such as leadership, communication, and critical thinking. Two internships, one focusing on scientific research and the other one on leadership-in-action, will allow scholars to apply their knowledge and leadership skills to concrete situations, and for the benefit of society. Prof. Kathryn Hess Bellwald, EPFL’s Associate Vice President for Student Affairs and Outreach who attended the event confirms that “the Laidlaw Program resonates with the School’s objective to form engineers and scientists who are not only technically excellent but also well-rounded citizens, conscious of their responsibility to contribute to society and well equipped to do so and aware of the necessity of being role models of respectful, ethical behavior.”

The program will provide them with a solid base for outstanding achievements while acting responsibly for the society and keeping community-centered values in mind.

The event also provided an occasion for a casual get together, where scholars got to know each other in a relaxed and festive mood. Whilst they all have their own views on what to expect of the Laidlaw program, they agree that they look forward the new discoveries and to this joint journey, focused on human connection. Within the next months, the scholars will dig into the first leadership workshop dedicated to self-awareness, teamwork and mindfulness, followed by a supervised summer internship in an EPFL laboratory to discover the world of research.

The discovery journey through the world of science and technology

To sensitize high school students for these issues and approaches, EPFL’s Education Outreach Department (SPE) mobilized nearly 50 EPFL scientists and students to offer a thematic day in collaboration with and for the high school Kantonsschule Solothurn (KSSO) within the high school’s premises, allowing approximately 600 pupils from a variety of study orientations to explore these links.

This thematic day inserts itself in the long-standing partnership SPE has with KSSO, aiming at strengthening ties for STEM promotion and outreach, in this case with a focus on highlighting links between STEM and sustainability.

Barriers to accessing STEM fields is certainly higher for none-technical oriented students. For this reason, SPE puts particular emphasis on a holistic approach with diversified and tailored activities for awareness raising, covering areas such as computer, life or basic sciences, engineering or architecture, through workshops, expert presentations, experiments as well as booths with interdisciplinary project presentations. Participants also had the opportunity to talk with students and scientists, and learn more about the research areas as well as how studies at EPFL look like.

Understand to make a better impact

Nathalie Morandini, Head of the Sustainable Development Division at the EPFL EssentialTech Center introduced the program with a keynote on how to harness science and technology to drive sustainable development, peace promotion and humanitarian action. Using the example of a fragile context such as unreliable power grids, lack of infrastructure or limited financial resources, she took the audience on a journey on how the EPFL EssentialTech Center collaborates with the many professors, researchers and labs across the School, showing how EPFL transfers knowledge. She encouraged the students to be curious, to look beyond the visible, to question how conventional technologies also can be applied to extraordinary contexts and create high value and quality solutions.

It is interesting to talk to young people and show them that whatever career they choose, they will be able to have a better impact on society if they understand early on the importance of being curious and open to other realities, cultures and ways of thinking.

Connecting the dots – or where the journey took us to

It is cool to recognize certain terms we learn at school in these research projects, and we can make a connection between school and research.

Clean drinking water, extraction and reutilization of C02 to fight climate change, adapting the construction industry to climatic challenges by using sustainable materials and constructing seismic-resistant buildings or even re-thinking energy, its storage and use… these are all topics that are directly of concern to the youth. Getting a taste of how science and technology can contribute to solving these problems has impressed the students and made them see the application of these concepts to real-world problems in a different light, as Fynn a third year scholar mentioned.

I find the program very interesting and it gives an insight into areas and applications we don’t even know yet. ​​​​​​

Two sides of the same coin: Jannik, a life sciences engineering student presented his biomedical student project which is more lab oriented and illustrated how introducing a given DNA sequence into bacteria in order to multiply it, is used to develop novel therapeutics, for instance for diabetes and prostate cancer research, thereby reducing to a certain degree animal experimentation. Julia on the other hand highlighted the engineering angle to biomedical research with her presentation of her project where she constructed an MRI-compatible ergometer to measure muscular flections, which is currently not possible in an MRI.

“With a scientific background I understand the science behind it. It was very well presented.

Impressed by what the EPFL-students have made

Two former KSSO-students, one studying Environmental Sciences and Engineering, the other Microengineering at EPFL were happy to return to their alma mater and reconnect as well as share their passion for science with their home crowd.

I find it impressive what the students have made out of their lives and what opportunities EPFL offers.

The day has not only raised the interest for these domains amongst potential future scientists and engineers, but also amongst the teachers who co-organized the event. A physics teacher was excited to see how EPFL brings these issues through connecting his students to people close to the actual projects, such as researchers or students, so as his pupils have a direct, hands-on immersion and are able to have peer-to-peer conversations. Equally important for him is the fact that it is a young generation reaching out to his scholars, which makes the connection more accessible.

“Young students, mixed genders, people who are close to the projects – that is what we need to spark the interest among our students.

Picture a doctor operating on a patient, and a nurse administering care. What gender did you attribute to each role? Even if you believe strongly in gender equality, chances are you pictured the doctor as a man and the nurse as a woman. That’s because we’re all subject to unconscious biases, or the implicit stereotypes and prejudices that influence how we view individuals and groups of people, including the characteristics we expect them to have. “We bathe in the society we live in, like dumplings in soup, and this gives rise to unconscious biases,” says Siara Isaac, a lecturer and scientist at EPFL’s Center for Learning Sciences (LEARN).

To make people more aware of their unconscious biases – and how these biases affect their social interactions – Isaac introduced a Micro-Ethics for Project Teams workshop three years ago. “The idea for the workshop came from a personal goal I had to better understand my own biases,” says Isaac, who also works as a teaching advisor at the Teaching support center (CAPE). “After speaking about the issue with Roland Tormey” – the head of CAPE– “we thought this kind of training would be useful at EPFL. The School supported our initiative. Our workshop is intended mainly for students in MAKE projects, but anyone can participate.”

First, spot the problem

The workshop is given every spring and fall semester and – based on a recent study appearing in Science and Engineering Ethics – has proven to be effective. Participants were surveyed a few months after completing the workshop, and 71% of them said they had become more aware of their unconscious biases, 84% felt they were better equipped to recognize prejudiced things they see and hear around them, 69% said they had changed some aspects of their thinking or behavior, and 62% reported using proactive strategies to make team discussions and decision-making fairer.

We bathe in the society we live in, like dumplings in soup, and this gives rise to unconscious biases.

“It was incredible to see the extent to which we react unconsciously in ways that don’t match our thoughts and beliefs,” says Anita Manzolini, an EPFL student who took part in the workshop. “These reactions stem from information and behaviors that have been forged by our culture and society. In the workshop, I learned that our initial reaction to something isn’t always consistent with what we believe in.”

An insidious issue

Unconscious biases are linked to the emotions that influence our decision-making and group dynamics. The biases are known to be one of the main reasons why there’s so little diversity in engineering programs at universities. Studies have shown these biases foster a discriminatory environment in which women and minorities have higher failure and dropout rates. “Our goal is to give students the tools to drive cultural change,” says Isaac.

It was incredible to see the extent to which we react unconsciously in ways that don’t match our thoughts and beliefs.

An EPFL survey on harassment, violence and discrimination conducted in 2021 found that 44% of female respondents had experienced inappropriate or derogatory comments, and nearly 25% had experienced unwanted physical contact (this figure rises to one-third for female students). “The problem with unconscious biases is you can’t see them,” says Isaac. “They’re very subtle and often run counter to what people say they believe in. They’re instilled in us, and they tend to show up when we’re stressed or have to make decisions quickly.”

The field of engineering isn’t immune to societal and ethical influences, even though it’s a science that tries to be as objective as possible.

So the first step is to draw people’s attention to their own biases. In Isaac’s workshop, participants are asked to complete an implicit association test that identifies their hidden stereotypes and prejudices. Then participants are taught strategies for working in teams in a more inclusive manner, and divided into groups where they can practice applying the strategies. An observer watches how the members interact and make decisions, along with any implicit stereotypes that emerge. The exercise ends with a group debriefing.

“The field of engineering isn’t immune to societal and ethical influences, even though it’s a science that tries to be as objective as possible,” says Andréa Montant, another EPFL student who took part in the workshop. “I learned that we should take the broader context into account when developing new technology, and that we should think about all the different kinds of people who may one day use it. Another factor to consider is project teamwork – how teams are managed and ways to address misunderstandings that arise between people from different backgrounds.”

Watch this video for an example of why it’s important to consider user diversity when developing new technology.

Make it awkward

At the end of the workshop, participants engage in play a role-playing game where they learn methods for responding to discriminatory comments and behavior. Isaac explains: “The game is based on things people told me they’ve experienced personally. I think everyone has either been the target of or witnessed inappropriate conduct but didn’t know how to respond on the spot. It’s not easy and takes practice. To help participants, I give them a set of cards describing many different ways to react.”

One way of reacting is to “make it awkward,” or to turn the tables by making an inappropriate comment embarrassing for the person who said it. This can be done by explaining your point of view, stating your values, affirming your boundaries or asking the person to think carefully about what he or she just said. “I’ve always been someone who doesn’t let such comments and behavior slide, whether they’re addressed to me or to someone else,” says Manzolini. “But what changed with the workshop is that I learned new methods. For example, I recently tried out one method – stating how a discriminatory comment made me feel – and it worked really well. I’d never done that before.”

Given that we’re all subject to unconscious biases, it’s also important to know what to do if you yourself say or do something unsuitable: you should recognize your mistake, take responsibility for it and state your intention to change. Sound easy? It is – but it may take some practice.

In 1597 during the Enlightenment, in his work Meditationes Sacrae, the British philosopher Francis Bacon coined the phrase ‘knowledge is power’. Four hundred years on, never has it been more important to have the knowledge and skills to address some of the core challenges faced by humanity today.

Education is key to gaining this knowledge and as we celebrate International Women’s Day, with a spotlight on education in the digital age, how can we ensure educational equality, free from the known bias contained within the machine learning algorithms that drive artificial intelligence?

PhD student Vinitra Swamy, in the Machine Learning for Education Laboratory led by Assistant Professor Tanja Käser, works on neural networks, pervasive as the general form of artificial intelligence models that we see everywhere and behind, for example, ChatGPT.

“One key objective of my research with machine learning is educational equality for all. When you look under the hood at what these models learn, you see human biases reflected from the data, leading to strong inequalities like man is to doctor as woman is to nurse. Mitigating this bias, especially at younger formative stages while students have developing minds and when these kinds of impressions last a long time, is so crucial,” she explained.

Following the explosion in digital education during the COVID-19 pandemic, Swamy is trying to increase course completion rates and address the problems of dropout for students undertaking Mass Online Open Courses (MOOCS), and when using education portals like Moodle, Coursera or EdX.

Specifically, Swamy and her colleagues have trained models to predict student success early in the course to make sure students get the most effective, unbiased, and personalized help they need to improve learning outcomes.

“We look at time series clickstreams of people interacting with education portals to make predictions. The patterns of students pausing or playing a video, when they post a question on the forum, when they answer a question on a quiz – all of these help us predict how to intervene for struggling students to get them back on track,” Swamy said.

“The key problem with using neural networks on human-centric data is that they don’t explain their decisions. The models can differentiate whether a picture is a cat or a dog very easily, but when asked to explain why it thinks the picture is a cat or a dog… that’s not inherent to the model design. It’s necessary to add attention layers at the front of the model or use post-hoc explainability methods to work out what models identify as meaningful.”

As the predictions generated by the models, and subsequent interventions, are all about helping students succeed, it’s important to understand why the models predict whether a student will pass or fail – whether the factors are completely arbitrary or not. Swamy used five state-of-the-art explainability models on top of the same neural networks, for the same students, and found that they all disagreed on the explanation for success or failure. This, she says, “is very concerning. If the methods you use to interpret the model are all giving different reasons for predictions of failure, which one can you trust?”

“If we know the model is making a prediction on student success based on demographic or socioeconomic attributes, that’s clearly worrying. However if all the explainers are systematically biased in saying what the models find important, that’s actually even more scary, because what you’re trusting to explain what the model is doing is biased also. Our initial research shows that there is pervasive disagreement in these explainers and our follow-on work aims to build trust around these explainers with human experts in the loop.”

All machine learning models come with sources of underlying bias and there are a lot of stages of the pipeline where this can enter: data selection bias, annotation bias, modeling bias, and downstream human bias. Swamy believes that for computer scientists it is critical to be aware of this issue and as well as how to address it.

“While I was at Microsoft AI, our team saw a lot of clients asking for toolkits to measure bias, choosing to use interpretable models, and understanding that responsible AI was important. In industry and research, we’ve seen a general uptick in interest for responsibility when it comes to creating these models and measuring and mitigating bias,” Swamy said.

For women and other minorities, Swamy believes machine learning for education can create a level playing field both in terms of educational and gender inequality and minority disparity, so long as attention is paid to all the places that the technology could go wrong.

“The education industry is primed for AI personalization, individualized homework assignments, auto-grading, and the kinds of techniques that we already see happening in the research world. Once those models make their way to the real world with real consequences, this issue of bias becomes a million times more important. It’s an ongoing issue.”