Public Discussion

  • Icon for: Brian Drayton

    Brian Drayton

    Co-Principal Investigator
    May 11, 2015 | 01:30 p.m.

    I presume that the ARTS system has to have some experience of me before applying its algorithms? I am thinking especially of the “spacing” process, which must have to deal with a lot of individual variation in tempo of learning?

  • Icon for: Everett Mettler

    Everett Mettler

    Co-Presenter
    May 11, 2015 | 08:20 p.m.

    Good question! One of the advantages of the ARTS system is that it reads your learning strengths automatically. It does this in real​ ​time, adjusting spacing in relation to the speed and accuracy of your responses. No prior experience is required for the algorithm to know your individual characteristics. The ARTS system is designed to seamlessly interpret and respond to individual variation and differences in the speed of learning for every individual learner, optimizing spacing and thus ensuring deeper and longer lasting learning. Hope that answers your question!

  • Icon for: Iliya Gutin

    Iliya Gutin

    Facilitator
    May 11, 2015 | 03:35 p.m.

    Great video! I was wondering at what “stage” in the learning process a student might encounter something like the PALMs software you have designed. Would a student use this as a kind of study-aid, or homework-module within their home or study environment? Or is this something that would actually be implemented during class or lab time, to augment the actual instruction.

    Thanks!

  • Icon for: Christine Massey

    Christine Massey

    Presenter
    May 12, 2015 | 07:39 p.m.

    Thanks for your question, Iliya — and we’re glad you like our video! Our PALMs have been used in a variety of ways. Typically, PALMs are intended to be a complement to other forms of instruction. In school settings, they are often used during class time, though they can be completed as homework as well. We often recommend that students use them in sessions of about 20-30 minutes at a time spread over multiple days. Because the PALMs are adaptive, different learners need different amounts of time to achieve full mastery levels. In class settings, the teacher (or other adult) has access through a teacher dashboard to a rich stream of real-time performance data for each student, so the teacher can monitor progress at the class level or the individual student level and adjust assignments accordingly. PALMs can also be used to efficiently refresh or review content learned previously, which really helps with long-term retention.

  • Icon for: Tony Streit

    Tony Streit

    Facilitator
    May 12, 2015 | 11:30 p.m.

    Very compelling piece – thanks! You do a great job at using the video format to tell the story and intent of your project. I’m impressed with what you are discovering about different learning styles. I’m curious though about what you’re learning about how disengaged learners struggle with STEM content. There’s much discussion in the field about how to best engage underrepresented young people and I wonder if your work can tell us anything about how best to tailor STEM learning to meet their needs. Thanks!

  • Icon for: Christine Massey

    Christine Massey

    Presenter
    May 13, 2015 | 05:04 p.m.

    Thanks for your kind words, Tony! I think we’ll have a better response to your question when we complete a large study that we’re running now that will give us data from more than 1500 students in urban schools, most of which serve under-resourced minority communities and have a large variety of students with different levels of performance and engagement. The software provides very detailed learning histories of each child, so we can look at their usage and progress with the software. We’ll also have a lot of co-variate data. I will say that one thing we’ve observed to date is that it seems to be a unique experience for some students in math when they see that they have completed all of the mastery levels in a PALM. We know that students who start at lower performance levels need more time to learn and practice, and the adaptive software makes sure they get what they need. This stands in contrast to much of their other classroom experience, where the curriculum marches on to a new topic before they feel secure with the old one, which leaves them thinking that they can’t learn math.

  • Icon for: Sarah Rand

    Sarah Rand

    Facilitator
    May 14, 2015 | 08:14 p.m.

    Love the format of the video- What program did you use to create it? It’s nice to have creative visuals with the audio.

  • Small default profile

    Dara Afraz

    Guest
    May 15, 2015 | 12:01 p.m.

    Thank you Sarah! This is Dara Afraz and I created this video. I’m glad you liked the format and I’d be happy to answer your question: I used Adobe Illustrator and Adobe Photoshop to prepare the vector graphics and used Adobe After Effects to create the animation. I hope that answers your question.

  • Small default profile

    Crispin Weston

    Guest
    May 15, 2015 | 03:15 a.m.

    Hello, I like the video and think this sort of sequencing tool is very important. My reservation is (a) with the apparent nature of the content you are using, which looks to me as if it incorporates only simple forms of interoperability, and (b) the implication that your sequencing engine – what your video is really about – seems to be hard-wired to your content. The critical requirement, in my view, before making real progress with adaptive learning is to ensure we can “decouple” the engine from the content so that you can sequence any content – and this requires open interoperability between the two, for launch, outcome reporting, and in metadata which allows the sequencer to understand the importance of the content in the structure of knowledge that you refer to. Such interoperability standards do not currently exist in satisfactory form – so your coupled content is probably necessary at the moment – but I’d be interested in your view of the importance of the of the future development of open interoperability standards for your work. Thanks.

  • Icon for: Philip Kellman

    Philip Kellman

    Co-Presenter
    May 15, 2015 | 11:02 a.m.

    Thanks for your comments, Crispin. I’m not sure I fully understand them all, but I can provide some information. Neither our perceptual learning technology nor our adaptive sequencing technology are hard-wired to any particular content. Also, they are compatible with a very wide range of display and response options; for example, we are currently applying them to surgical simulation and training in a project sponsored by the US Office of Naval Research and driver training in a project sponsored by AAA Foundation. Part of the beauty of these techniques is their almost universal applicability. Of course, we work closely with domain experts to develop the learning goals and content in areas such as mathematics and science learning, medical learning and surgical training, aviation training, driving and pharmaceutical discovery. But the basic technologies apply readily to any domain. You can find out more from published work in K-12 mathematics and geography (Bufford et al, Proc. Cog. Sci. Society, 2014; Kellman et al, TopiCS in Cog. Sci., 2010; Mettler et al, 2011, Proc. of the Cog. Sci. Society), perceptual learning of biological kinds (Mettler & Kellman, Vision Research, 2014), and medical learning in areas such as dermatology (Rimoin et al, J. Am. Academy of Dermatology, 2015) and histopathology (Krasne et al, J. Pathology Informatics, 2013). Hope that helps.

  • Icon for: Kathy Perkins

    Kathy Perkins

    Director
    May 15, 2015 | 11:01 p.m.

    Nice video. It really helped explain your project. You mention that the same algorithm has been used across various content contexts. It seems that for some content reaction time is really useful measure — when you are trying to build that automaticity of knowledge (like sight words in reading or multiplication tables). Is there particular categories of content you feel this algorithm is better suited for helping learners learn?
    Also, are the PALMs you’ve developed available for anyone to use? Thanks.

  • Further posting is closed as the showcase has ended.

  1. Christine Massey
  2. http://www.ircs.upenn.edu/pennlincs/html/index.html
  3. Director of Research & Education, Institute for Research in Cognitive Science
  4. Adaptive Sequencing and Perceptual Learning Technologies in Mathematics and Science
  5. http://www.ircs.upenn.edu/pennlincs/html/current.html
  6. University of Pennsylvania
  1. Philip Kellman
  2. https://www.psych.ucla.edu/faculty/page/kellman
  3. Distinguished Professor, Psychology Department
  4. Adaptive Sequencing and Perceptual Learning Technologies in Mathematics and Science
  5. http://www.ircs.upenn.edu/pennlincs/html/current.html
  6. University of California, Los Angeles
  1. Everett Mettler
  2. http://kellmanlab.psych.ucla.edu/
  3. Postdoctoral Researcher
  4. Adaptive Sequencing and Perceptual Learning Technologies in Mathematics and Science
  5. http://www.ircs.upenn.edu/pennlincs/html/current.html
  6. University of California, Los Angeles
  1. Rachel Older
  2. http://kellmanlab.psych.ucla.edu/
  3. Lab Manager
  4. Adaptive Sequencing and Perceptual Learning Technologies in Mathematics and Science
  5. http://www.ircs.upenn.edu/pennlincs/html/current.html
  6. University of California, Los Angeles

Adaptive Sequencing and Perceptual Learning Technologies in Mathematics and Science
NSF Award #: 1109228

This research and development collaboration between labs at UCLA and the University of Pennsylvania creates and studies STEM learning technology that integrates principles of perceptual and adaptive learning based on well-established research in cognitive science to help students become more effective and efficient learners. Perceptual learning accelerates learners’ abilities to recognize and discriminate key structures and relations, helping students develop expert-like abilities to recognize and extract important information in complex science and math representations, such as models of molecules in chemistry and algebraic equations in math. Adaptive learning algorithms use a constant stream of individual performance data, combined with principles of learning and memory, to adapt the learning process to each individual. These algorithms, known as the ARTS system (Adaptive Response Time-based Sequencing), continuously track the speed and accuracy of a learner’s responses to different types of problems and use the learner’s own data to optimize how items should be spaced and sequenced in during learning. The software also uses the student’s data to guide him or her to objective mastery criteria. The methods and algorithms we study are currently being instantiated in web-based Perceptual and Adaptive Learning Modules (PALMs) in areas including K-12 mathematics, high school and community college chemistry, and medical education.