Using Open Recognition to Map Real-World Skills and Attributes

Part 1: From Folksonomy to Taxonomy

Update: also check out Part 2: Building the workflow and platform!

CC BY-ND Visual Thinkery for WAO

Open Badges exist to democratise the means of credentialing: anyone can issue a badge for anything. The trouble is that it can be difficult to do so, because people tend to describe things using folksonomies: “Oh yeah, they’re a good listener”

This article expands on our recent posts about endorsement, offering a method for aligning Open Recognition with skills taxonomies. For example, “good listener” is reframed as “active listening”.

The approach we outline helps you gather useful information by talking to people you know, both personally and professionally. It also helps you use an AI assistant to do some of the laborious work and do what machines do best: pattern-matching.

Here’s the 10-step process:

1. Come up with questions for your contacts and put them in an online form
2. Share the online form with relevant people
3. Organise the answers you get in a way that’s easy to understand
4. Ask an AI assistant to review the answers and give you a summary
5. Instruct the AI assistant to review skills taxonomies
6. Compare the skills you’ve identified with taxonomies to see how they match up
7. Summarise your skills using a table
8. Come up with the metadata for Open Badges based on your skills (coming in Part 2!)
9. Create badges using an online platform (coming in Part 2!)
10. Ask contacts to endorse your badges (coming in Part 2!)

In our experiment, we gathered 20 responses within a 24-hour time period by explaining the process via social media posts. We’ll use this example to bring the process to life.

We’ll cover steps one through seven in this post and save steps eight to ten, which deal with turning skills into badges, for Part 2.

1. Come up with questions for your contacts and put them in an online form

The simpler the better. For this experiment, we started off with one contextual question which used checkboxes. We followed this with three diagnostic questions that allowed for free text entry.

In what context do you know this person?

Tick all that apply

Professional — we have worked together directly

Professional — we have interacted but not worked together

Professional — other

Personal — I am a family member

Personal — I am a friend

Personal — other

What do they know a lot about?

This could be a subject area (e.g. ‘Geography’) or something more specific (e.g. ‘Identifying rare books’). Feel free to go into detail.

What are they particularly skilled in?

This could be a hard skill (e.g. ‘coding in Python’) or a human skill (e.g. ‘facilitation’). Feel free to give examples and context.

What behaviours do they exhibit which you, or others, find useful?

This is a catch-all for everything else that makes this person different/valuable to others. You might want to talk about their impact on the work you’ve done, or something you’ve observed that doesn’t fit elsewhere.

We used Google Forms to collect answers. Make sure the form is visually appealing and straightforward. Specify who should complete it, like people who know you well, and include links to any privacy policies.

Based on our experience, people are more willing to fill out forms if they’re anonymous, and this approach tends to yield better feedback.

2. Share the online form with relevant people

LinkedIn is a good platform for sharing the survey, but we also posted on the Fediverse via our Mastodon account.

You might want to email the form to certain people or groups as well. If you’re using LinkedIn, consider tagging individuals you’d like to hear from, especially those you’ve worked with recently.

3. Organise the answers you get in a way that’s easy to understand

Google Forms has a handy way to export to a spreadsheet:

Go to ‘File / Download’ and generate a PDF which can be referenced by an AI assistant such as ChatGPT.

Ensure that you’ve turned word wrapping on so that none of your data gets hidden.

4. Ask an AI assistant to review the answers and give you a summary

You can view the conversation I had with ChatGPT 4 at this link. However, we’ll break it down and give you the prompts and if you want to use them yourself.

Give some context and upload your PDF

We used the AskYourPDF plugin for ChatGPT. Here is the prompt we used:

I’m going to share a PDF with you, so I’ll need an upload link. The contents of a PDF are responses to an online form. Here is the introductory text:

— — —

Open Recognition in your network [EXPERIMENT]

👋 Hello! Thanks for filling out this form, which should take less than 5 minutes. This will form an example for an upcoming article showcasing how Open Recognition can work in practice, mapped against skills taxonomies. We do not capture your name or identifiable details, so any words you enter will not be attributable to you.

Please complete this form as honestly as possible. Your results will be captured using this Google Form (privacy policy), combined with those from other people, and fed into ChatGPT (privacy policy) to map against skills taxonomies. WAO may also store the data on other platforms, in accordance with our own privacy policy.

We appreciate you taking part. This example is for Doug Belshaw. If you know and have interacted with Doug in a personal or professional context, please continue to fill out this form. If not, thanks for your interest but please do not continue. All questions are optional. 🙂

— — —

Once you’ve given me the upload link and I’ve given you the doc_id, we’re going to analyse the responses to the questions to:

1. Synthesise results

2. Discover patterns

3. Match attributes against skills taxonomies

After you input this text, you should get a link to the AskYourPDF website, where you’ll be asked for the document ID of the PDF you upload. Just copy/paste it into the chat window.

Check the summary

AI assistants sometimes get things wrong, so check the summary ChatGPT generates. Ours looked like this:

You could ask ChatGPT for deeper analysis — for example, segmenting what your personal contacts said about you, compared with what your professional contacts indicated.

5. Instruct the AI assistant to review skills taxonomies

To compare skills, we need a reference list. In this experiment, we used a review of skills taxonomies on the UK government website. A direct link to the relevant PDF can be found here. It references the O*NET (USA), ESCO (EU), and Nesta (UK) skills taxonomies.

To upload the PDF, do as you did earlier: use ChatGPT to get a link that lets you upload the document through the AskYourPDF plugin. Here is the prompt we used:

I need another PDF upload link. This is from the GOV.UK website and is an “Independent report. Review of skills taxonomies: May 2022. This paper, commissioned by the Skills and Productivity Board, assesses the usefulness of existing skills taxonomies in relation to the work of the board.”

ChatGPT then summarises the paper, explaining what the different skills taxonomies are useful for.

We’re interested in all of the skills taxonomies, as O*NET focuses on skills at a higher level than ESCO and Nesta.

6. Compare the skills you’ve identified with taxonomies to see how they match up

ChatGPT can now see data from two PDFs. The first has feedback from your contacts about your strengths. The second contains data about skills taxonomies. We can now use this information to match what people say you’re good at with lists of skills used by different organisations and platforms.

Here’s the prompt we used:

So now that we’ve got some examples of skills taxonomies, I’d like you to take what we learned about Doug Belshaw from the online form and match it against a list of work-related skills.

List them all in bullet-point form with a justification for each.

ChatGPT produced the following:

Some of these are obvious, but others are perhaps less so. It’s useful to have the specific terminology to use to describe your skills.

7. Summarise your skills using a table

Having the list from the previous step is helpful, but it’s even better to compare the data in a table format. Here’s the prompt we used to do that:

Now create a table from this data in a way that I can copy/paste

ChatGPT produced the following:

While we’re more than just words in boxes, having a structured way to be able to talk about our skills with people and systems is incredibly impactful.

Listing your ‘top 10 skills’ just became a whole lot easier.

Conclusion

Completing this process is highly valuable for understanding the skills your network attributes to you. However, sometimes people struggle with the confidence to reach out even to people they know and trust. So, ideally, the above workflow would consist of automated prompts for you and messages to your contacts.

The second post in this series will outline ways in which you can convert these skills into Open Badges. By doing this, you can showcase skills using any platform using skills taxonomies that are readable by both humans and machines.

If you use this approach, please share your results with us and your network! We’d love to see how this works for others.