Peeragogical Pathways

Briefly describe your organisation or individual main areas of work

Can we create learning opportunities leading to meaningful work for billions of people? A place to start is with sites like Stack Exchange, GitHub, and Wikipedia. For many, these are places to ask questions, share answers, solve problems, and understand the world we live in better. And yet many others are not presently able make much use of these resources.

Since 2012, the Peeragogy Project has been building a collection of informal design patterns for effective peer learning and peer production. In the AI for Good programme, the new company Hyperreal Enterprises will translate these patterns into a software tool that can be used to organise open source materials into "peeragogical pathways" that support learning and development. In the words of their originator, Christopher Alexander:Man Tried in the Court of the Lion

“Each pattern describes a problem which occurs over and over again in our environment, and then describes the core of the solution to that problem, in a way that you can use this solution a million times over, without ever doing it the same way twice.”

We are concerned with problems of human infrastructure. Our solutions will augment existing resources and workflows with AI methods that support learning and co-production at scale.

Please outline up to 2 examples of relevant professional experience (if applicable).

This proposal builds on prior interdisciplinary collaboration in the Peeragogy Project and formal research into applications of AI within collaborations.

The Peeragogy Handbook (4th ed. forthcoming 2020) includes a growing collection of design patterns that model the activities that take place in peer learning and peer production. We have innovated on the classic pattern language template by adding a “What’s Next” step to each of our patterns. Thus, each of our patterns is linked with a number of “issues” that reflect the way that the pattern is being put into practice in the project. The same methodology can be used to make sense of the issues and challenges faced in other projects.

The other strand of research examines collaboration with an eye to building “mixed initiative” AI systems. This work has resulted in publications in the journals Artificial Intelligence and Argumentation [1], [2]. We modelled the way people collaborate to create mathematical proofs, by building a “dialogue game” that describes exploratory mathematical thinking, and constructed an empirical model of real-world mathematical dialogues. This work was additionally supported by small-scale experiments with a novel reasoning system [3].

[1] http://bit.ly/36wPkQx [2] http://bit.ly/302Gnfn [3] http://bit.ly/2N97L6b

What social issue does your project address(e.g. health, climate etc.)?

Skills development, including “reskilling” for the digital economy, through participation in open source projects. This will have further knock-on benefits. GitHub’s director of social impact, Admas Kanyagia, observes that open source has the “power to drive collaboration and shared expertise to drive innovation and change and to build communities particularly with multiple and diverse perspectives" [1].

[1] http://bit.ly/2FunB71

Provide a short description of the social issue you are trying to solve. How have you identified this?

The central issue we are tackling is the “skills gap”, within the broader context of a transition to a knowledge economy. The problem has many well-documented facets. Hired.com found a significant gap in the UK’s technical skills [1]. Nesta reckons there will be an increasing need for social skills and learning skills as well [2]. Nevertheless, the Learning and Work Institute highlights that participation in learning within the UK is at an all-time low [3]. Furthermore, the OECD found that simply deploying technology within classrooms does little to bridge the gap between advantaged and disadvantaged students [4].

We will focus on inclusive uses of technology to support skill acquisition at scale. We are inspired by the London-based Founders and Coders collective [5] which is able to provide free technical training to selected applicants, leading to gainful employment. We want to expand the benefits of peer learning to anyone who has an internet connection. We will build on prior research into peer learning and peer production, and emphasise interacting systems, diverse skillsets, and human relationships.

[1] https://hrd.cm/2QxleXs [2] http://bit.ly/2sXP3aM [3] http://bit.ly/303DKty [4] http://bit.ly/35FMebS [5] http://bit.ly/35BVltO

At what stage is your project?

Within the Peeragogy Project, our design patterns for peer learning and peer production exist at a ‘proof of concept’ stage. In particular, we have used the patterns within this project to help manage our workflow. In the new project outlined here, we will move this material into a software ‘prototype’ that will interoperate with existing sources of open data.

Describe your project - what is it you are doing?

The core technical activities combine human learning, machine learning, and logic:

1. Harvest issues from GitHub, Stack Exchange, and Wikimedia projects.

2. Tag a subset of this material using the Peeragogy design patterns (e.g., Newcomer, Heartbeat, Wrapper, etc.). Expand the collection of patterns to ensure meaningful coverage.

3. Tagging will give us a collection of terms that can be used to detect patterns in a broader data collection using a standard machine learning approach.

4. Operationalise the design patterns in code, with multiple implementations. We will be able to detect patterns when they appear, and also implement them programmatically when they do not. For example, we might detect that a project has a low level of ongoing activity but no "Heartbeat", and accordingly suggest a regular meeting time.

5. We will orchestrate the patterns in combination with each other, discovering complex synergies that lead to learning success, and forging new connections between related projects so that they can support each other.

6. We will be able to demonstrate new potential team-ups and other opportunities for exchange; new workflows and value propositions, learning prerequisites and outcomes.

We will engage with relevant designers and users at least two workshops.

How do you think your project will help solve the issue you have identified? How does your use of AI differ from what is currently being used within thisfocus area?

AI applications to human learning are often geared towards students in formal educational settings. Automated tutoring is poised to become a major business, and several large-scale investments and deployments already exist. However, this does little to assist with informal learning and online collaboration; and, per the OECD report mentioned above, questions remain about the social impact of such interventions.

Design patterns have been applied on an experimental basis within computational design projects, starting from a vision outlined by TP Moran [1]. The very first wiki was created to keep track of design patterns related to programming. However, there has been surprisingly little attention to the use of design patterns to bridge between collective and artificial intelligence.

This project will draw on symbolic methods and will combine them with “big data” to progressively build a model of the patterns in open source projects, including Wikimedia projects and Stack Exchange question-and-answer dialogues [2], [3]. ‘Bots’ are well-accepted in these environments, however a focus on learning support is novel.

[1] TP Moran. “(Artificial, intelligent) architecture: Computers in design”, Architectural Record 149 (1971), 129–134.

[2] Wikimedia issue tracker: https://phabricator.wikimedia.org/

[3] Stack Exchange question and answer network: https://stackexchange.com/

How will your involvement in AI for Good - the funding, curated programme and profile building - help realise the potential of your project?

The AI for Good initiative will help Hyperreal Enterprises develop into a new kind of company. We will be able to connect with local talent and other UK-based initiatives during the programme. At the same time, we must recognise that the “skills gap” is properly global in scope [1], and we aim to address it on a global scale. Hyperreal Enterprises will build on public data to enhance the existing global knowledge commons. In the curated programme we will co-learn in direct contact with other progressive initiatives, and contact related endeavours sponsored by the Centre for Collective Intelligence Design [2]. We will be able to give something back by contributing to a better understanding of how CI projects can succeed. More specifically, while there are many ways to apply AI within CI [3], the present project moves in the direction of building computational models of CI itself. We aim to show this is an appropriate and effective way to scaffold engagement with CI, and skill development more broadly. Alongside working software, we will use workshops and hands-on learning experiences to co-develop an engaging vision.

[1] https://nyti.ms/2Qxvlf0 [2] http://bit.ly/35t8Lsc [3] http://bit.ly/2R3DPtw

What gaps or challenges will you need to overcome in order to develop this project?

Lack of interest is the biggest documented obstacle to learning [1]. Many open source projects and new wikis fail to attract contributors, and even successful projects typically decline in popularity as time goes by [2]. AI can help by working transversally across projects and assembling communities of collaborators, users, and supporters.

Addressing this challenge requires a different way of thinking about software that takes into account its social side. Joe Corneli and colleagues at the University of Edinburgh developed an innovative data science course and ‘data fair’ model [3], [4]. Building a similar initiative that cuts across institutional boundaries will be a novel challenge.

The primary strategy to address these challenges will be through transdisciplinary collaboration. For example, whereas both Free Software and Wikipedia have well-documented problems with gender representation, our patterns describe domain general collaboration strategies, and can help work against insular trends to help build a truly inclusive movement. For example, we can connect communities by sharing a wiki page or software issue with potentially interested stakeholders who are unfamiliar with the medium, while scaffolding these “Newcomers’” engagement.

[1] http://bit.ly/303DKty [2] http://bit.ly/35wf9Pj [3] http://bit.ly/2QAHkZc [4] http://bit.ly/35wPZjt

Who are the individual team members working on the project? What are their roles and key skills?

Joseph Corneli, PhD (Director of Hyperreal Enterprises, Ltd., based in Musselburgh) will be involved .5FTE throughout the March-September programme.

Joe’s doctoral thesis (2014) examined “peer produced peer learning” in the context of PlanetMath.org, an innovative online learning community for mathematics. He has since completed two postdocs in “computational creativity” and “social machines”. This proposal combines these research themes into a practical software system.

Other collaborators within the Peeragogy project will be involved on a volunteer basis. A “mid-term” report will be submitted to a workshop associated with the International Conference on Functional Programming, to be co-written with long-term collaborator Raymond Puzio (PlanetMath.org, Ltd.).

Joe has previously written about patterns in two papers published in the Pattern Languages of Programs conference series: “Patterns of Peeragogy” (2015) and “Patterns of Design” (2017). Joe also brings a good awareness of the AI space, for example our current project will use methods similar to recent work by Google Research [1] and NIST [2], and will allow us to move in the direction of understandable computer code and AI for programming [3].

[1] https://github.com/google-research-datasets/coarse-discourse

[2] Scientific Statement Classification over arXiv.org, https://arxiv.org/pdf/1908.10993.pdf

[3] Semantic Representation of Data Science Programs, IJCAI.

What wider support or capacity do you have to deliver your project? This could include institutions, colleagues and formal and informal partnerships?

I have links to Lorna Campbell on the Board of Wikimedia UK, to Jeremy Knox who co-directs the University of Edinburgh’s Centre for Research in Digital Education, and to Neil Mulholland in the Edinburgh College of Art. Mulholland’s 2019 book “Reimagining the Art School: Paragogy and Artistic Learning” published with Palgrave builds on ideas that I developed in my thesis, and provides links to other creative communities across the UK. Mulholland has invited me to participate in a creative arts workshop that will explore peer learning and peer production in a Scottish context. Knox has offered to help arrange demos and other knowledge exchange and impact activities. Wikimedia UK may be able to support an additional workshop.

I have contacted Co-operative Development Scotland to ask for assistance organizing Hyperreal Enterprises as a co-op, and will liase with the Co-operative University to understand how we can contribute to their developing initiative.

In parallel with this submission, I have developed a course called “Peeragogical Innovations” that I hope to teach in Tufts University’s Experimental College this Autumn. This would be an opportunity to expand on the themes of this proposal internationally, by engaging students and invited speakers in further rounds of codesign.