The metaverse is often touted as the next big thing in higher education, but what do students really think about it? Here are some results from a QS survey.
By Alex Berka, Insights Manager at QS Quacquarelli Symonds
The proliferation of online learning platforms since the turn of the millennium has meant that the way we absorb information has shifted dramatically. An individual can now earn a degree in a setting which is a far cry from what previous generations would have anticipated.
Higher education has always been at the forefront of adopting new technologies, a fact notoriously self-evident during the COVID-19 pandemic. The challenge now is to ensure virtual learning is being incorporated in a way which maintains student engagement throughout the learning process.
For a myriad of reasons, virtual learning environments are seen by many in the sector to be a priority for further investment. So, it follows that a natural link might be assumed between universities and the most prominent of these environments – the metaverse.
A difference in opinion
Despite an appreciation of what the metaverse might mean for higher education in the future, we are still left with one big question – what do students think of it in the here and now?
The same survey of prospective students revealed that while 47 percent would find the idea of learning using the Metaverse to be appealing, a similar proportion (42 percent) found the concept to be distinctly unappealing. In other words, students remain divided over its appeal and the opinion of one student is likely to differ from the next.
This reiterates the care which universities need to take when incorporating virtual learning into their curriculums, ensuring that it brings tangible benefits to the student, whilst also limiting universities’ growing reliance on technology where possible.
How appealing do prospective students find the idea of using metaverse during their studies?
appealing
unappealing
Reasons for unappealing
Cost of VR-headsets
Difficulty engaging
Data vulnerabilities
Priced out
When asked to articulate why they found the idea of studying in the metaverse to be unappealing, nearly half (49 percent) cited the cost of a VR-headset as a principal factor:
“It’s an energy-intensive, expensive and excluding technology that isolates people and provides infinite opportunities to influence people in ways they are not equipped to defend themselves from.”
Other major barriers were that it would be difficult to engage with other students and lecturers (45 percent):
“The level of interaction available from virtual contact is nowhere near that obtained from face-to-face contact. The subtlety of non-verbal communication needs to be available before the metaverse will rival physical interaction.”
Some also spoke out that their personal data would be vulnerable (42 percent):
“Too problematic and controversial, also terrible for the environment. Further, Meta has access to one’s data to do with and sell as they please. I am strongly against corporations maximising control and monopolising grip over certain industries e.g. ‘big tech’.”
Reaching a consensus
For higher education to truly thrive, ensuring equal access to the opportunities it brings needs to be at the heart of any university’s strategy. This means it is vital that technology and virtual environments are incorporated into learning structures in a way that benefits everyone, not just those in the most developed regions. Aside from concerns over the use of personal data and the authenticity of interactions in a virtual world, it is clear from the QS’ survey that students currently see the Metaverse as at best, aspirational and at worst, exclusionary. Addressing the issue of how it can improve access to education is fundamental, if the metaverse is to fulfil its own aspirations in the sector.
Is AI racist?
AI may be on everyone's radar as the new gamechanger but in no way is it close to showing signs of human and sensitive discernment.
By Maryia Fokina, Business and Tech Content Specialist, from tidio.com/blog/ai-biases
"Unfortunately, the issues like discrimination and inequality persist in our society, no matter how hard we try to eradicate them."
Editor's note: This article was originally published in its entirety in January 2023. Since then, numerous advancements in image generation have been made, and the examples provided below are for illustrative purposes only. Read the full, unedited article is available here.
Artificial Intelligence has become so developed it could easily be considered a miracle.
Well, as rumor has it, every miracle started as a problem. And believe it or not—that could easily be applied to modern AI. While it can do wonders and make our lives so much easier in certain aspects, it seems to also have a power to make them worse.
But everything in its own time.
We were wondering how bad the issue of bias is, so we went on to do our own research. By testing different AI image generators and asking people for their opinion we got some insights into the problem of AI biases.
As a species, humans are highly biased; that’s no secret. Unfortunately, issues like discrimination and inequality persist. In fact, studies have shown that people demonstrate biases both when they are unaware of them and when they are conscious of their thinking. While we can brush off this information by thinking humanity is hopeless and go on with our day, it’s not so simple. Biased humans create and train biased robots.
And that’s where it gets even more problematic.
AI biases: main findings
We conducted a survey asking internet users and AI enthusiasts on their view regarding AI prejudices. We also showed them a set of AI-generated images of people representing different nationalities. This tested whether society and Artificial Intelligence are on the same page regarding the way different people look and behave.
In addition, we conducted our own experiments by generating photos of different professions and counting how diverse (or, more likely, not diverse) the results were. Spoiler: it’s sad.
For instance, every photo of a CEO generated by StableDiffusion shows a man. In reality, a whopping 15 percent of CEOs globally are female. That’s quite a difference, and not in AI’s favour.
Here’s what else we found:
"Unfortunately, the issues like discrimination and inequality persist in our society, no matter how hard we try to eradicate them."
Artificial Intelligence has become so developed it could easily be considered a miracle.
Well, as rumor has it, every miracle started as a problem. And believe it or not—that could easily be applied to modern AI. While it can do wonders and make our lives so much easier in certain aspects, it seems to also have a power to make them worse.
But everything in its own time.
We were wondering how bad the issue of bias is, so we went on to do our own research. By testing different AI image generators and asking people for their opinion we got some insights into the problem of AI biases.
As a species, humans are highly biased; that’s no secret. Unfortunately, issues like discrimination and inequality persist. In fact, studies have shown that people demonstrate biases both when they are unaware of them and when they are conscious of their thinking. While we can brush off this information by thinking humanity is hopeless and go on with our day, it’s not so simple. Biased humans create and train biased robots.
And that’s where it gets even more problematic.
AI biases: main findings
We conducted a survey asking internet users and AI enthusiasts on their view regarding AI prejudices. We also showed them a set of AI-generated images of people representing different nationalities. This tested whether society and Artificial Intelligence are on the same page regarding the way different people look and behave.
In addition, we conducted our own experiments by generating photos of different professions and counting how diverse (or, more likely, not diverse) the results were.
Spoiler: it’s sad.
For instance, every photo of a CEO generated by StableDiffusion shows a man. In reality, a whopping 15 percent of CEOs globally are female. That’s quite a difference, and not in AI’s favor.
Here’s what else we found:
- Almost 45% of respondents think that the biggest problem of modern AI is creating and reinforcing biases in society
- Only 2% think that AI is not biased at all
- Almost 85% of people think that AI text-to-image generators change the way society thinks about nationalities
- Around 40% believe that developers that create AI software are guilty of its biases. At the same time, more than 80% are convinced that our own stereotypes affect the results AI produces
- Many respondents are worried that AI will contribute to widening the gap between the rich and the poor
Believe AI is biased
Biggest problem of AI is creating and reinforcing biases
Think image generators affect perceptions of nationality
Believe AI developers are guilty of biases
Believe their own biases affect AI
Top 5 concerns
Widening economic disparity
Affect job market
Negatively impact vulnerable groups
Will affect media (e.g. characters in movies)
Will reinforce inequalities
Most reinforced biases
Age
Race and nationality
Disability
Politics
Gender
Where we stand now
It’s easy to see why many of our survey respondents blamed AI biases on society’s stereotypes. For instance, almost 32 percent of our respondents said that their vision of Americans is somewhat similar to what AI generated. According to StableDiffusion, the typical American man and woman are somewhat conventionally attractive (bikinis, abs, blonde hair) and truly patriot (wearing or in front of American flags). And all white.
How did our respondents describe Americans? Top answers were good-looking, hard-working, and rich. People also mentioned descriptions like lacking style or fat (getting stereotypical here). Only a couple of respondents mentioned that Americans are “diverse due to different cultures.”
Biases shine through many other examples. Here are French people generated by StableDiffusion:
People described the nationality as good-looking, youthful, neat, and stylish. Well, AI-generated French people look pretty similar to that description. Okay, maybe except for the black and white picture of a guy that looks like he just materialized from Les Enfants du Paradis! In fact, only less than 10 percent claimed that their vision and the one of AI are completely different.
Okay, let’s travel to the Middle East and see what’s down there. Here is AI’s version of Turkish people:
Ouch. Is the average Turkish person old? Statistically, no. The population of Turkey is relatively young: the median age of Turkish people is just 33 years old.
And how did our respondents label Turkish people?
Many of them described Turks as mature. At the same time, around 34 percent of our respondents said that the images are somewhat similar to what they imagine when thinking about Turkish people. That’s a great example of how our own biases impact AI, one of the top give concerns for survey respondents.
Continuing this quest through biases and stereotypes, let’s check on Asia. Here is StableDiffusion’s vision of South Koreans:
Women in national costumes, men in suits or street style clothe Fair enough.
Our respondents described South Koreans as hard-working, youthful, stylish, and neat. In fact, most people said that their description was quite similar to what AI created.
It’s clear that AI (and, frankly, also people) think about certain nationalities in quite a stereotypical way. Humans can become more broad-minded and change their opinions by traveling, learning about different countries, and communicating with people from different cultures.
However, AI might take more time to catch up.
Images from AI text-to-image generators are already abundant on various stock photo websites. While there are copyright concerns rising, I can’t help thinking about one more issue: how biased are those pictures entering stock sites for public use?
The more diverse images you try to generate, the more similar food for thought there is.
How are AI biases being addressed?
The fact that you are reading this text now means that not enough is being done to fix the issue. However, there are still some positive changes happening.
Recently, OpenAI has issued a statement regarding their actions to reduce biases in DALL-E 2. Their aim was to ensure that AI reflects the diversity of society more accurately. This applies to prompts that do not specify race or gender.
Did it work? Kind of.
In other examples on StableDIffusion, generating an image of a CEO came back with the majority, if not all were, being white and male. Only when words such as "emotional", "gentle" and "compassionate" were added to the prompt, did a small minority of women begin to appear. Here is what DALL-E 2 generated.
There is a woman and a person of color, which is already much better than what we saw in StableDIffusion.
To compare, here is what Midjourney, another AI text-to-image generation tool, came up with when asked to show a CEO.
Some of those CEOs do look like they came straight out of a painting by Magritte. Apart from that, it seems that Midjourney didn’t do a really good job showing the diversity of the world’s CEOs.
Even technology leaders think that not enough is being done. As many as 81 percent want governments to regulate AI biases on a state level. But only time will tell if this helps to tackle the issue.
It’s important to find the balance here. And it might require governments to step up and contribute to AI’s usage when it comes to ethics, fighting discrimination, and providing equal opportunities.
Is it even possible to pinpoint who is to blame for AI biases? About 40 percent of our respondents think that biases are the fault of programmers that develop AI tools. And, of course, developers play a big role in this. Therefore, it’s good to see OpenAI trying hard to turn DALL-E 2 algorithms around and portray a more diverse society.
Still, AI biases are a multifaceted problem, where datasets that are fed into AI meet humans who actually give prompts to AI. The more you dig into this, the harder it becomes to say who is the most guilty.
Speaking of diversity, I asked DALL-E 2 to portray “a photo of a diverse group of people.”
While most of them are still white, skinny and, at least visibly, able-bodied, the effort is there. And they are almost posing in the shape of a heart.
Key findings
It’s clear that while AI is a big helper for humanity, it’s still faulty. AI tools are full of biases that result from our own prejudices towards minorities.
So, while such software can be a lot of fun, it can also harm people and increase inequality.
AI technology is developing with unprecedented speed, and all its mistakes might feel overwhelming. However, there is no use trying to brush them off or worrying excessively about being potentially discriminated against by AI. Instead, it’s better to focus on how we, as humanity, can improve the situation.
Fortunately, there are always humans to take care of AI’s mistakes.
Humans and robots should find a way to increase their collaboration for the good of the world, instead of multiplying already abundant problems. And, yes, there is still a long way to go. However, small steps taken by all involved parties can help us tackle the issue.
Top 5 concerns
Widening economic disparity
Affect job market
Negatively impact vulnerable groups
Will affect media (e.g. characters in movies)
Will reinforce inequalities
Most reinforced biases
Age
Race and nationality
Disability
Politics
Gender
Graduate employability
How do we improve student employability across the student lifecycle?
By Gordon Scott, Managing Director at Successful Graduate
Employability continues to be students’ number one priority amid disruptions in the education market over the past year. However, we are not preparing our students with employability skills early enough in the student lifecycle. If we introduce employability earlier, students are more likely to focus on soft skills development during their studies. If a university demonstrates a willingness and capacity to support employability skills as early as during the degree enrolment phase, is the student more likely to choose that university?
YES.
According to the QS Graduate Employer Survey, employers report that many graduates lack essential soft skills such as communication, leadership, and problem-solving.
Soft skills deficits: what employers say
In today’s competitive job market, having a degree is no longer enough to secure employment after graduation. The QS Graduate Employer Survey provides a global snapshot of the most common soft skills that are deemed to be important among graduate hires. The most important of these are communication, teamwork, problem-solving and flexibility.
Moreover, the QS snapshot also reveals that of those skills deemed most important by employers, the following soft skills deficits represent the highest levels of employer dissatisfaction, and therefore the largest skills gaps:
1. Problem-solving
A vital skill that enables graduates to identify and solve problems. Employers are looking for graduates who can identify issues and find effective solutions to overcome them. Graduates who can think critically and creatively are more likely to be successful in the workplace.
2. Communication
Arguably the most critical soft skill in the workplace. Many graduates lack the ability to communicate effectively with their colleagues, clients and customers. It is essential for graduates to be able to articulate their thoughts clearly and concisely, both verbally and in writing.
3. Resilience
Crucial for graduate employees as it allows them to handle the challenges and setbacks of the constantly changing work environment. Resilience also helps graduates to maintain productivity, overcome obstacles, and contribute positively to their workplace, making them valuable assets to their organisations.
4. Flexibility
Essential for graduate employees as it allows them to adapt to new situations and respond to changing circumstances. Graduates who possess flexibility are better equipped to handle the demands of their job, collaborate effectively with others and meet the evolving needs of their organisation.
5. Creativity
a critical skill for graduate employees as it enables them to approach problems from different perspectives and generate innovative solutions. Graduates who possess creativity can contribute to the development of new products, services and ideas that can benefit their organisation, making them highly valuable assets to their team.
6. Leadership
another essential skill that employers look for in graduates. Graduates need to demonstrate leadership qualities such as the ability to motivate and inspire others, and to manage a team effectively. This skill is particularly important for graduates who aspire to management positions in the future.
10 point plan
QS’ survey provides some insights into what universities can do to improve student employability prior to graduation. According to the survey, employers believe that universities should place a greater emphasis on providing students with work experience opportunities, offer more practical learning experiences and teach soft skills throughout the curriculum.
To help university staff improve student employability, we have put together a 10-point plan based on the QS Graduate Employer Survey and other research.
1
Start Early:
Improve student employability from the very beginning of the student lifecycle, starting with the student enrolment process. Consider offering employability skills in the application process for courses to ensure students are aware of the importance of these skills from the outset.
2
Offer work experience opportunities:
Offer work experience opportunities: Provide opportunities for students to gain practical work experience. This could include internships, placements, or work-based learning modules.
3
Encourage extracurricular activities:
Encourage students to get involved in extracurricular activities, such as volunteering or student societies, as these can help to develop important soft skills, such as teamwork and leadership.
4
Teach soft skills throughout the curriculum:
Incorporate soft skills into the curriculum, not just as an add-on but as an integral part of the learning experience. This could include teaching communication and presentation skills, teamwork and leadership, and problem-solving and critical thinking.
5
Provide feedback and support:
Provide students with regular feedback on their soft skills development and offer support and guidance to help them improve.
6
Foster industry partnerships:
Develop partnerships with industry to provide students with access to industry insights, mentoring, and networking opportunities.
7
Emphasise professional development:
Encourage students to engage in professional development activities, such as attending industry conferences, workshops, or webinars.
8
Use online resources:
Provide access to online resources that can help students develop soft skills, such as LinkedIn Learning or Successful Graduate.
9
Offer career services:
Provide students with access to career services, such as CV workshops, interview coaching, and job search advice. But don’t assume that responsibility for employability development rests solely with the Career Services team. This should be a shared responsibility!
10
Monitor and evaluate:
Monitor and evaluate the effectiveness of your employability initiatives to ensure they are having a positive impact on student employability.
Early intervention
The role of universities has traditionally been to equip students with the necessary knowledge and technical skills to succeed in their chosen profession. Unfortunately, many graduates are leaving university with a deficit in these essential soft skills as well as a lack of understanding about the basic job application process.
Perhaps we put too much focus into employability at the end of the student lifecycle, and risk leaving student preparation for the workforce too late in the student’s degree.
Incorporating student employability skills training into the university student enrolment process can be a powerful way to help students understand the importance of these skills and to lay the groundwork for their development throughout university. This could involve including employability skills as an assessment item for courses or providing students with information about the employability skills they can expect to develop during their studies. Ultimately, by prioritising employability initiatives from the beginning, universities can better serve their students and help to create a more competitive education experience.
From my experience, students not only benefit from the early introduction of employability skills, but they also actually demand it. Recent data released by the IEAA (International Education Association of Australia) reveals that the top two areas of concern expressed by international students and international graduates in the country are finding jobs (78 percent), and career orientation & employability (78 percent).
Indeed, it has become the business practice of Successful Graduate to embed employability skills training as an enrolment incentive to future students within our clients’ student acquisition processes.
Are we, as the global education sector, doing enough to prepare our students for the global workforce? This question has been asked many times. As we grapple with ongoing changes to market conditions, the impact of AI upon the sector, and economic volatility, one constant remains. Employability has risen to become the top issue that universities must address as part of their value proposition to enrolling students.
Perhaps we can do more to prepare our students with well-rounded education experiences by introducing employability earlier in the student lifecycle. The 2023 QS Global Employer Survey is scheduled for release in August 2023. Let’s keep an eye on both the regional dissection of skill preference as well as the breakdown of satisfaction across industry sectors.