Challenges for educators when using AI

Should we even use artificial intelligence in teaching and learning?

People are concerned about the hastiness of AI, particularly chatGPT, which is gaining momentum in the education space and the world we live in quicker than the line for hot savouries at a staff PD day. From shopping list and meal plan recommendations to fitness goals, personal playlists, lesson plans, and even a personal favourite of mine, using AI to show your significant other you have more empathy than you have in real life (no link required, this is through empirical research!), chatGPT can easily slip into almost every element of our lives. Does it seem like we have much say in it at the moment - not really, unless you want to completely bury your head in the sand and pretend it's not happening. 

As of the 1st of June 2023, less than 10% of schools and universities have formal guidance on AI. So where does that leave us as educators? Do we embrace it and use it to enhance teaching and learning, or do we just pretend that it will go away if we ignore it (much like Microsoft Internet Explorer/Edge - disclaimer: this is my new favourite browser with the integration of Bing being able to analyse web pages in browser - never say never!)

Proceeding with caution will clearly be the key. We have little individual control over how the tools are developed or shared, but we can decide how we interact with them. We must also recognise that we have been using AI for a long time (spam email filters, mobile phone photography, Alexa, Google Maps, all social media, Amazon shopping recommendations, Music and podcast apps) and have already contributed to its development. However, we can be mindful of how we use the tools, especially when working with young people.

At this time, with very limited guardrails on AI, I think that these are the areas where we need to build awareness and understanding; in no hierarchical order, just how they came into my head:

1) AI literacy

This is for us as learners, teachers, and the learners we work with. We must role model a high level of AI literacy to support those we work with. AI will transform our lives. We must understand how to interact with tools thoughtfully and ethically so they can emphasise our best human traits and support us in areas we need assistance. (Image below was created by AI with the prompt “We must understand how to interact with tools thoughtfully and ethically so they can emphasise our best human traits and support us in areas we need assistance in a simple style”!)

2) Privacy and information sharing

How much are you prepared to give up to AI to get what you want? What happens to your data when you enter it into ChatGPT? Do you know what Bing does with your search terms? Do your school policies and procedures allow you to enter school data and student work into AI? As we become more comfortable with the idea of AI, we will share more with it to perhaps get more back. However, your data can be exploited by AI, especially if you enter personal and confidential data - this can be sold or even used for identity theft or fraud. Advancements in AI can also use biometric data (for example, how we unlock our phones with facial recognition or a thumbprint) or voice and location data to monitor our movements and breach our privacy. If you have used an online tool to track your genetic origins, this data could be used to discriminate you from services such as healthcare, insurance and finance in the future due to genetics. This impacts you and any of your relations born after you - a heavy price to pay to find out that you are 3% Celtic. (Image below created by AI with the prompt “keeping data safe online in a simplistic style”.)

3) Data sovereignty 

This is related to the previous point; however more specific in that understanding that knowledge held by some people, specifically in New Zealand, is a taonga. I am not the right person to share expertise on this, but from my understanding, AI does not have the right to tell stories that belong to Iwi and Hapū; context is so important, but once AI takes hold of this precious mātauranga it can rewrite it and tell it any way that it likes. Who decides what information goes into LLM, and how can we make sure that some knowledge remains sacred?

4) Digital divide

It is relatively free to use AI now, as long as you don't mind contributing your data to developing more powerful AI. However, even in a matter of months, there seems to be an endless plethora of tools that fit our exact needs, from sourcing and reading literature to creating bespoke videos that are all built on the openAI API but have that extra layer of usability. Almost all of these apps are freemium, meaning you can use them in a limited capacity. Currently, we can all access the 3.5 version of chatGPT, while people on the paid version can use the more advanced 4.0. How long is it until openAi stops updating the public open-access version? As other apps bring this in, it could well be a ''pay for'' service, and in the end, the level of AI support you can access will depend on your income, again creating a divide. It will be like the difference between being able to access a personal assistant and having to do it all alone. Making the growing education gap even more chasmic.

5) Bias and stereotype reinforcement

A colleague of mine, Sophia Zhang, gave a great talk at a recent AI in education forum held by academeyEX; Sophia shared that there is a significant gender gap: Only 16 % of all AI professionals are women. UNESCO's Director-General Audrey Azoulay states, "There is an urgent need to rebalance the situation for women in AI to avoid biased analyses and to build technologies that take into account the expectations and needs of all of humanity"". Human bias is built into AI through conscious and unconscious assumptions about race, gender, or other ideological concepts. While AI designers are attempting to avoid this bias by understanding more about the training data, creating diverse teams and enabling effective feedback loops, there is no quick fix that does not involve cleaning the entire dataset for AI. You can check for bias yourself, and a good learning activity would be to use an image creator to type in terms and discuss the potential bias with your students. The image below was created by Bing image creator with the simple prompt: Doctor. Why is one of the images just a drawing, while the others attempt to depict real people?

6) Reducing creativity and personal voice

There are two trains of thought here; if used proactively, it can amplify your creativity by using elements of AI to build on an idea or thought. However, it can be used lazily to create without stretching the imagination too far by crafting prompts that will give you something similar to what you hoped for without going through the process. For example, Microsoft is soon releasing their own design tool that will suggest integrating images, backgrounds and even hashtags. What will this do to our creative desires? I'm not sure at this stage, as it seems like it is human nature to take shortcuts, but it is also human nature to create, so we may have the chance to be more creative in an offline way. You could even build on your skills to be more empathetic to the people you love (note to self). 

This is only a short list, and there are many other challenges, such as hallucinations and false information, filter bubbles created through AI and on a larger scale, the cost to run AI, the energy sources used to power AI, and of course, the world'sworld's end is also a concern for some very influential people.

The most straightforward and common-sense approach to the challenges we face in education has been presented by UNESCO, built on the  Beijing Consensus. Reinforcing a human approach may be the only way to survive an unknown future alongside AI.

(This post was written by a human - as with all posts on AI, it will probably be out of date in about a week. In fact, I rewrote this a couple of times based on the rapid developments in AI! The video at the end was created by adding the text from the blogpost into Lumen5 which is linked earlier.)