Tag Archives: algorithms

Open the pod bay doors…

People are not fearful of algorithms, they’re fearful of agendas that the algorithms represent.

2001-a-space-odyssey

Over the last few weeks I have been discussing and listening to people’s views on the intelligent campus.

One topic which has resulted in a fair bit of controversy is the concept of using artificial intelligence to support teaching and learning. This isn’t some kind of HAL 9000 computer running the campus and refusing to open the library doors until Dave the learner has finished their essay. This is more about a campus system being able to learn from the users, take that data, do some analysis and make suggestions to the user on potential ideas for improvement and useful interventions.

Imagine a learner arriving at campus with the intention of writing an essay, needing a quiet place in which to do this. They check their Campus App on their smartphone and it recommends a location based on the ambient noise levels and the type of environment the learner has used before. It could take into account the distance from the coffee shop, depending on if coffee is used as a distraction or supports the learner in writing their essay. The learner can of course ignore all this and just go to where they want to, the app provides informed guidance and learns as the learner does more learning activities and which spaces they use.

Another scenario, is a teacher planning a session, with some relatively interactive and engaging learning activities. They ask the intelligent campus where is the best place for this to happen! The system takes on board the preferences of the teacher, the availability of rooms, information from previously successful similar sessions and any feedback from learners. The teacher can then make an informed choice about the best space for this session. After the learning, the system asks for feedback so that it can learn from and improve the decisions it makes.

I think some of the issues (or should we call them problems and fears) that people have with a concept such as this is they feel any such algorithm is secret and hidden and will have a built in bias.

hal-9000-reflecting-daves-entry-in-stanley-kubricks-2001-a-space-odyssey

As I wrote in my previous blog post on open analytics I said

So if we are to use algorithms to manage the next generation of learning environments, the intelligent campus, support for future apprenticeships and data driven learning gains, how can we ensure that we recognise that there is bias? If we do recognise the bias, how do we mitigate it? Using biased algorithms, can we make people aware that any results have a bias, and what it might mean?

People are not fearful of algorithms, they’re fearful of agendas that the algorithms represent. But if we make these algorithms open and accessible, we could mitigate some of those concerns.

So if we are to use algorithms to support teaching and learning, could we, by making the algorithms open, ensure that, we remove some of those fears and frustrations people have with a data approach? By making the algorithms open could we ensure that staff and learners could see how they work, how they learn and why they produce the results that do?

This does bring up another issue that people have mentioned which is the bias any algorithm has, bias which comes from the people who write it, sometimes consciously, sometimes unconsciously. There is an assumption that these algorithms are static and unchanging and written by people who have an agenda. As we know from using Google and other algorithms, these are constantly changing and being tweaked.

Could we go one step further and allow people to edit or even create their own algorithms? Allowing them to make suggestions on how they could be improved, creating new analytics that could benefit the wider community.

We need to embrace the benefits of a smart campus, because the technology is already here, but we need one which learns from the people who use it; we need to ensure that those people are the ones who inform and guide the development of that learning. They are able to and can decide which intelligent campus decisions to benefit from and which they can ignore. By making the whole process open and accessible, we can provide confidence in the decision making, we can feel we can trust those decisions. We mustn’t forget that giving them that literacy in this area, is perhaps the most important thing of all.

Remember that in both scenarios above the learner and teacher both, ultimately, have the decision to ignore the intelligent campus decisions, they can decide themselves to close the pod bay doors.

Opening the algorithms: Could we use open analytics?

globe

Do you remember when the Google algorithm wasn’t that good, well it was good, but today it’s better!

Many years ago if you searched for a hotel on Google, so you could find out if there was car parking, or to find the website for the restaurant menu, the search results most of the time were not the hotel website, but hotel booking sites offering cheap hotel rooms. Pointless if you already had a room, and all you wanted to know if you had to pay for car parking, or what time you could check out. The problem was that the hotel booking sites worked out how the Google search algorithm ranked sites and “gamed” Google search.

Today, the experience is very different, the algorithm usually results in the actual hotel website being the top hit on any search for a specific hotel.

Google had worked on the algorithm and ensured what they saw as the correct search result was the one that was at the top.

One thing that many people don’t realise was that Google not only worked on the software behind the algorithm, but that they also use human intervention to check that the algorithm was providing the search results they thought it should be. If you wonder why Google search is better than search functions on your intranet and the VLE this is probably why, Google use people to improve search results. Google uses people to both write the algorithms and to tweak the search results. Using people can result in bias.

laptop

So if we are to use algorithms to manage the next generation of learning environments, the intelligent campus, support for future apprenticeships and data driven learning gains, how can we ensure that we recognise that there is bias? If we do recognise the bias, how do we mitigate it? Using biased algorithms, can we make people aware that any results have a bias, and what it might mean? If we are to, like Google, use human intervention, how is that managed?

The one aspect of Google’s search algorithm that some people find frustrating is that the whole process is secret and closed. No one, apart from the engineers at Google really knows how the algorithms were written and how they work, and what level of human intervention there is.

So if we are to use algorithms to support teaching and learning, could we, by making the algorithms open, ensure that, we remove some of those fears and frustrations people have with a data approach? By making the algorithims open could we ensure that staff and learners could see how they work, how they learn and why they produce the results that do?

Could we go one step further and allow people to edit or even create their own algorithms? Allowing them to make suggestions on how they could be improved, creating new analytics that could benefit the wider community.

Is it time for open analytics?

Thank you to Lawrie Phipps for the conversations we had after the Digital Pedagogy Lab: Prince Edward Island conference and this blog post.