AI Writing

So I saw some tech-guru on Twitter pontificating about how awesome AI writing assists were and I thought, let’s give that try. So, I signed up for a demo of a product, and it’s pretty slick. (HASHTAG NOT AN AD #notsponsored). The thing that I used is called Moonbeam which is aimed at helping you become a better writer (and yes, it does help with the blank page syndrome).

My initial though is: slick, but not quite there yet for niche interests (which holds out hope for teaching (more on that in a minute).

I gave it the prompt “The Disasterous Choice to Teach Using an LMS”.

It then provided me with a fleshed-out draft – three arguments against, three arguments for and a conclusion. I edited the three arguments for out and generated this:

The Disasterous Choice to Teach Using an LMS

Introduction: Discuss the reasons why some instructors choose to use an LMS, or learning management system, to teach their students and how this negatively impacts learning.

Some instructors choose to use an LMS because it is a convenient way to deliver course content and track student progress. While this may be appealing at first, using an LMS can have negative consequences for learning. For example, a more passive learning environment can create difficulty for students in accessing and understanding the material. Additionally, using an LMS can limit instructor creativity and flexibility in designing and delivering instruction. As a result, the choice to use an LMS should be based on what will best meet the needs of the instructor and the students.

There are several potential disadvantages to using an LMS, including the following:

There are several potential disadvantages to using an LMS, including the following:

Increased potential for cheating: With an LMS, it is easier for students to find and copy answers from each other, which can lead to decreased grades and a loss of learning.

Limited interaction with instructors: If problems arise or clarification is needed, students may have difficulty getting in touch with their instructors in a timely manner through an LMS.

Technical difficulties: An LMS can be prone to glitches and crashes, which can interfere with classes and disrupt learning.

High cost: Some LMSs can be expensive for schools or individual teachers to implement and maintain.

a) It can be difficult to keep track of all the different features and functions of an LMS.

It can be difficult for educators to keep track of all the different features and functions of an LMS. This can lead to frustration and confusion for both teachers and students. An LMS can also be expensive, and many school districts cannot afford to purchase one. There is also a lack of research on the effectiveness of using an LMS in the classroom. Some educators believe that an LMS can actually hinder learning rather than promote it.

b) An LMS can be time-consuming to set up and maintain.

An LMS can be time-consuming to set up and maintain, which can take away from time that could be spent teaching.

An LMS can be difficult to use, which can lead to frustration for both teachers and students.

An LMS can be expensive, which can limit its use to only those who can afford it.

An LMS can be inflexible, which can make it difficult to adapt to changing needs or requirements.

c) Some students may find it difficult to navigate an LMS due to design of the course or the LMS itself.

Some students may find it difficult to navigate an LMS due to its design or the LMS itself, leading to frustration and a feeling of being overwhelmed. This can impact their ability to learn effectively.

An LMS that is designed with user-friendliness in mind can help alleviate some of these issues. Additionally, providing support and training for students on how to use an LMS can also be beneficial.

Ultimately, it is important to consider the needs of all students when choosing to use an LMS for teaching purposes.

Lack of control over the learning environment, dependence on technology, and limited interaction with students are some of the challenges that arise when using an LMS for instruction. These issues require significant investment of time and resources if they are going to be overcome.

Ineffective for certain types of learners is another potential downside of using an LMS for teaching purposes; however, this can be remedied by taking into account the specific needs of each student group before designing a course content strategy around an LMS platform

Now that’s not a bad starting point. I could flesh that out and expand on the points. And that’s the value that Moonbeam proposes – giving folks an option to edit all the way through (which likely improves their AI model). If I were generating content as a content provider, I definitely would be using this as a way to flesh out ideas and keep things a little more focused while working rapidly.

Now if I were a niche content producer (oh say, talking about educational technologies) then the downside to this is you’re not providing much additional value over a Google search as I’m sure the AI is using scraped data for one of it’s source, and so the niche stuff which doesn’t have a lot of linkage, or doesn’t have a lot of major traction, like most of the educational technology focused blogs, then it probably won’t deliver anything but a surface understanding of the subject. I can’t imagine the wrongness of something about punk (actually I can, because I generated a blog post about how 1982 is the best year for punk – it included the sentence: 1984 was a crucial year for punk music, as it saw the rise of crossover poppunk and hardcore metal.)

The testimonials on the front page of the site, well, they tell you that (if you’re an educator) your students are already using the tool. A quick social media search shows that is true as well. If it’s on Twitter, chances are there’s thousands more on private Discord servers and SnapChat.

The hope for education? Well, it’s the same argument we had about the web destroying education – access to all the information in the world does not provide the context one needs to form a deep understanding of the subject. Instead of rehashed, crappy essays (and please, can we get away from essays) and going towards more participatory ways of demonstrating learning I’m sure higher education will simply try to outlaw these things and adhere to stricter more surveillancy kinds of approaches to solving the problem. The problem is that more participatory ways of demonstrating learning is that they aren’t easily scalable – or are they? One thing that MOOCs taught me is that with a large enough bank of assessments, it is totally possible to demonstrate learning at scale (now assessing… that’s a different matter). Look at the DS106 Assignment Bank. It’s got clear assignments linked to loose themes (which could be your learning outcomes if that’s how you roll) that students can select from. I suspect that these would be difficult to do through AI. Another strategy might be to use the AI and show how you’d improve that output as part of the essay building process.

Will we get there though? The labour costs of doing the DS106 assignment bank is huge – and the assessment of 600+ students (thinking about first year folks) would be considerable. Maybe there’s ways around it, I’m sure that people could come up with clever ways to assess at scale – but anything offloading assessment to students or peers is a band-aid on the way higher education is running.

ETEC 511: Tipping Point, A Critical Case Study Proposal

One of the observations I have made over the years, and particularly over the last decade at McMaster University, is that the LMS has mostly displaced the use of personal websites for teaching. The reasons for this are multifaceted and contextual to individual institutions, however, at McMaster, I have observed that it is most likely related to the course themes of Digital Labour (once in the LMS, it is easier, and less labourious to keep using the LMS and the labour of using the LMS can be offloaded to teaching assistants) and Attention (student preference is to have all learning in one place). However, there are secondary contributing factors, which would fall under Algorithms (enhanced ability to track and observe course activity) and Sustainability (not from an environmental standpoint, but a course sustainability practice). There is also a factor of culturalization – since 2011, LMS use has not been mandaked. The makeup of faculty has skewed younger and with that pre-LMS teaching has faded from institutional memory. In many cases, no one even thinks that teaching outside institutional systems is possible or even desirable.

ETEC 511: Project Proposal

Checking UDL and Accessibility: A Checklist for Educators.

Most countries have adopted disability legislation that requires legally compliant interventions at the organizational level, including businesses, services, and educators, to accommodate the broadest possible range of human experience (Doyle, 2020).  We are proposing a Universal Design for Learning (UDL) checklist tool intended to help adult educators in any context examine learning materials that they develop to address accessibility and neurodivergent users. 

This tool will function in a branching decision tree to help educators determine if their learning objects have the affordances they need to accommodate all learners. The decision points will be supported with an explanation of how to employ UDL principles for that criteria and a rationale for why it is essential to support that factor, including some examples of specific (dis)abilities the factor supports.

Intended Users

This tool is designed for educators (designers, teachers, trainers, consultants) who create learning materials (digital or physical) for adult learners.

Setting Adult Learners Up for Success

We aim to enable adult educators to create learning materials that accommodate all facets of UDL, so the educational experience can be inclusive and accommodating to all learners.

One in five people has a neuro-difference, such as dyslexia, ADHD, autism, or anxiety disorders (Doyle, 2020). According to Statistics Canada, in a 2017 study, 14% of Canadians aged 25 to 64 with disabilities reported having at least a university qualification, compared with 27% of those without disabilities (Statistics Canada, 2019). These disabilities included neurodevelopmental condition(s) (NDC), a mental health condition (MHC), or both. These are diagnosed conditions, so the number of undiagnosed people is likely far higher. Those adults with undiagnosed learning (dis)abilities have multiple barriers to learning in place and often must look for accommodations in the workplace or post-secondary education. In fact, 43% of employees with (dis)abilities (and neuro-differences) do not feel comfortable approaching their employer to ask for accommodation (Business Disability Forum, 2020). We want to create a tool that supports educators when creating learning materials for all learners, regardless of ability, and mitigating the need for after-the-fact learning accommodations.

There are increasing resources for UDL principles; however, educators need to learn how to apply them in an organized, straightforward manner to determine whether their learning object/module/course meets UDL guidelines.

“The UDL framework is grounded in three principles:

●      Multiple means of representation – using various methods to present information and provide a range of means to support.

●      Multiple means of action and expression – providing learners with alternative ways to act skilfully and demonstrate what they know.

●  Multiple means of engagement – tapping into learners’ interests by offering choices of content and tools; motivating learners by offering adjustable levels of challenge.” (TEAL Center Staff, n.d.)

Our checklist will help educators accommodate the three UDL principles. Other resources list what UDL considers but do not provide the depth or breadth required for someone to design learning materials with all the principles in mind. Given that UDL is a newer approach, many educators are unfamiliar with its benefits and drawbacks. There is a lack of resources that provide this sort of support and an easy-to-access and use tool that provides guidance and awareness on the full range of learner abilities, particularly those with unrealized and unacknowledged learning disabilities. This tool will be helpful for new instructional designers and educators to acclimate them to UDL principles.

Creating an Interactive Decision Tree

We believe Twine will allow us to create branches for different aspects of UDL, be accessible to all educators and be easy to use. If there is an overlap between resources or rationale, we can link to the same resources without having to create them twice. Twine allows for the inclusion of videos and links to web resources, which allows us to address the multimodal framing of UDL within our design.

Our tool configures its users by determining what can be checked. Our project group selects what is essential from a UDL lens to be incorporated into learning materials, focusing on adults with unacknowledged learning disabilities. We aim to provide educators and designers with an easy-to-use tool that helps all adult learners, regardless of barriers or learning abilities.

UDL Checklist Items

In our tool considerations, we would like to include factors such as:

●  What devices might learners be using and what are their technical capabilities?

●  Have you considered different (dis)ability needs?

●  By making it accessible for one group, are you making it less accessible for others?

●  Have you considered inclusivity (language, font style and size, colour palette, etc.)?

●  Does the approach have an alternative method of delivery?

●  Does it provide appropriate and sufficient cognitive support (organizing clues, background information, scaffolding)?

●  Does this approach comply with ACR standards? Are there additional provincial compliance requirements?

●  Does the mode of assessment consider necessities and fairness (time constraints, presentation method, learner option, authenticity, supporting resources)?

Using Inclusion to Necessitate Usability

This tool is a checklist. One of the limiting factors of checklists is that we will have to determine what is included in the list and how that list is sequenced. Woolgar (1996) would consider this an element of configuring the user. This limitation makes the tool more usable and less overwhelming for practitioners new to UDL principles. Gee (2005) states that we learn best when we understand how things fit into a larger meaningful whole.  By reducing the broad scope of UDL to more manageable subtopics, it becomes easier to digest and learn about the benefits of UDL. Chunks have long been proposed as a basic organizational unit for human memory (Laird et al., 1984). Chunking is an essential strategy for learning complex subjects. Laird et al. (1984) demonstrate that a practice mechanism based on chunking can speed up task performance and may be capable of leading to more exciting forms of learning than just simply improving the speed of acquisition.

The output of Twine creates accessible objects delivered on the web. It requires low bandwidth and does not require significant computer processing power to operate, allowing people to access this tool from a desktop or mobile device regardless of location or operating bandwidth.

Determining Usability and Success

Using the usability specifications (as defined by Issa and Isaiah, p. 34) we will examine performance measures such as the responsiveness of the completed site (is there a significant delay from clicking on an item?) and preference measures from a target audience focus group to examine whether this checklist tool improves their understanding of UDL. If time allowed, a longitudinal study could be conducted to see how the covered UDL principles were recalled one year after the user introduced the checklist.

  

References

Are Mental Health and Neurodevelopmental Conditions Barriers to Postsecondary Access? (2019, February 19). Retrieved October 9, 2022, from https://www150.statcan.gc.ca/n1/pub/11f0019m/11f0019m2019005-eng.htm

Business Disability Forum. (2020). The great big workplace adjustment survey: Exploring the experience and outcomes of workplace adjustments in 2019-20. Retrieved October 13, 2022, from https://businessdisabilityforum.org.uk/policy/the-great-big-workplace-adjustments-survey-2019-20/

Doyle, N. (2020). Neurodiversity at work: a biopsychosocial model and the impact on working adults. British Medical Bulletin. 135(1), 108-125. https://doi.org/10.1093/bmb/ldaa021

Gee, J. P. (2005). Learning by design: Good video games as learning machines. E-learning and Digital Media, 2(1), 5-16. https://doi.org/10.2304/elea.2005.2.1.5

Initiative, W. W. A. (n.d.). WCAG 2 Overview. Web Accessibility Initiative (WAI). Retrieved October 15, 2022, from https://www.w3.org/WAI/standards-guidelines/wcag/

Issa, T., & Isaias, P. (2015). Usability and Human Computer Interaction (HCI). Sustainable Design, 19–36. https://doi.org/10.1007/978-1-4471-6753-2_2

Laird, J. E., Rosenbloom, P. S., & Newell, A. (1984, August). Towards Chunking as a General Learning Mechanism. In AAAI (pp. 188-192).

TEAL Center Staff (n.d.). TEAL Center Fact Sheet No. 2: Fact Sheet: Universal Design for Learning. LINCS. Retrieved from https://lincs.ed.gov/state-resources/federal-initiatives/teal/guide/udl 

Woolgar, S. (1990, May). Configuring the User: The Case of Usability Trials. The Sociological Review, 38(1_suppl), 58–99. https://doi.org/10.1111/j.1467-954x.1990.tb03349.x

ETEC 511 – IP #3: Algorithms

Option I: Content Prioritization

“At a time when state funding for public goods such as universities, schools, libraries, archives, and other important memory institutions is in decline in the US, private corporations are providing products, services and financing on their behalf. With these trade-offs comes an exercising of greater control over the information, which is deeply consequential for those already systematically oppressed…” (Noble, p. 123)

Think and respond to the following questions:

  • Explain in your own words what “content prioritization” (Noble, p. 156) means (give some examples) and how (in lay terms) content prioritization algorithms work. With control over the “largest digital repository in the world” (Noble, p. 187), how have Google’s content prioritization algorithms been “consequential for those already systematically oppressed”? How do they impact your professional life? (give specific examples and briefly discuss)
  • What are some ways PageRank impacts your personal life? (specific examples and briefly discuss) (How) can you impact PageRank? Explain.

Content prioritization essentially is a resorting algorithm based on a myriad of factors. In Google Search, it is based on things like location data, prior search history, demographic information about your account, and other personalization. While this might seem like a good and useful thing, it does often lead one down a rabbit hole. Google Search, like most sites, wants your attention. The more time you spend with it the better. The more searches, means that they can build a better profile of you and what you want to see. I will come back to that in a moment.

When people started figuring out how to improve their own sites search ranking, they started to manipulate link text to manipulate Google’s algorithm for search ranking. This lead to using potential misleading link text (the early internet history’s version of being Rick-Rolled) to mislead users through a process called Google Bombing. As these became passed around early social media, they also caused Google to rank them higher in priority based on the number of searches being run for the search term. The one that might be memorable was during the second Iraq War, anti-war groups made an effort to link “miserable failure” to the White House’s website. Typically, these Google Bombs were not long-lasting, as you can see from Google Search trends for the phrase “miserable failure”. However, their impact was.

The manipulation of search ranking was seen as a strategy from radical right groups (McSwiney in Devries, Bessant & Watts, 2021, p. 25) to access, recruit and radicalize users. The sheer volume of racist propaganda online is almost pervasive. If one of Google’s search algorithm key ranking criteria is based on volume of links, it is no wonder that racist, biased sites get pushed up the rankings. One of Noble’s arguments throughout the book is that while Google builds the algorithm that pushes certain sites to the top of the pile, they are not responsible for it (Noble, 2018). So, it follows when Google autocompletes a search query with a stereotypical response, that has an impact on the viewer – either reinforcing a negative view or potentially introducing self-doubt and the ranking algorithm every time it is clicked.

In my professional work, I often am searching websites for documentation about educational technology products. I am often working on a work account which has little to no demographic information, never search logged in, with no location technology able to be queried. Essentially, my work account is a bit like a burner account. So typically, no, I do not see any evidence of discrimination, however documentation does have discrimination built into it – the types of archetypes used, the images of people describe more about the company than many think. However, from a search perspective, I do not use Google autocomplete ever, I do not use Google as a sole search provider, I move around (Duck Duck Go and Bing are two suitable competitors), so that there isn’t much to give to one provider.

Back to the popularity contest that is PageRank, and attention. While I do not use Google exclusively, PageRank’s algorithm strategy is pretty pervasive amongst search. It is part of what was taught at Udacity’s big “build a search engine” MOOC (I completed this in 2012?) and is what Yandex and Bing use to scrape the web for links, and to count the number of links that point to a site with a set of keywords. It is a common strategy and would like yield common results – except you do not have the ranking algorithm component, but both ranking and link scraping work hand in hand. The first way that PageRank influences daily life is the reliance on what is popular over what is factual. I have seen this over and over, popular misconceptions – and how tales take over the reality of what happened. Sure, we have context for some of that (in that disadvantaged groups often do not get their stories told at all) but Google’s focus on popularity assumed (when PageRank was developed initially) that people are mostly truthful. Instead, 15 years later, it is less about people and more about how many resources can be deployed to increase site ranking, essentially privileging the wealthy (who are disproportionally white, heterosexual and male) and resourceful.

Well, one way to circumvent PageRank is to Google Bomb it to uselessness. Essentially fill it with obfuscated information. However, that only punishes you – because it makes it less useful to you. One other way to make PageRank be influential is to pay less attention to it and SIFT the sources. Go straight to the source rather than Google Search everything (the number of times I have seen someone search for the website, rather than just type the site address is more than I can count in the last decade). Use other search engines. Avoid giving attention to things that are false. Support platforms that do not combine personal data with search results.

References

Devries, M., Bessant, J., & Watts, R. (Eds.). (2021). Rise of the far right : Technologies of recruitment and mobilization. Rowman & Littlefield Publishers.

Noble, S. (2018). Algorithms of Oppression : How Search Engines Reinforce Racism. NYU Press.