Workflow: A Recommender System for Thinking and Insight Generation

InfraNodus has a built-in insight recommender system, which can be used to help you generate insight from any text or data. It is available via the Insight Panel at the bottom of the screen and it will provide you advice on every step of your interaction with the graph:

It will recommend which topics you can explore deeper or connect in a new way. You can follow the instructions (e.g. "think of a relation between your physical practice and using a mirror to provide real-time feedback on movement") or use the built-in GPT-3 AI module to generate ideas, facts, and research questions automatically based on the insight that InfraNodus provides. 

Below is a step-by-step guide and a video that explains how it works.

How it Works: Ecological Thinking through Cognitive Variability

When you start adding the data, InfraNodus Insight system will automatically analyze its network structure. If the data network is dispersed, our system will point out the different ways you can bridge the gaps in your ideation process and emphasize the main topics. If the data network is homogeneous, the system will recommend the different ways to develop the periphery and diversify the ideas in your discourse.

The insight recommender system is programmed to maintain ecology in your ideation process, guiding you through the thinking process so that a long-term dynamic equilibrium can be maintained between a diverse set of ideas. You are encouraged to combine synthetic and disruptive thinking, alternating between the specifics and the whole:

The model shown above on the Figure 1 is related to panarchy, which embraces the notion of non-equilibrium stability and a natural process of development based on the acceptance of growth and crisis as the essential parts of any dynamic process. A similar approach is used in cognitive behavior therapy where there is a general acceptance that a period of evolution (e.g. building up a positive behavior) may be followed by a relapse (going back to the old habits), which, in fact, allows us to get feedback on what works and what doesn't and enter the evolutionary cycle again in an empowered manner. As you can see, this approach is popular in many different contexts and we think it can be very beneficial to use it in the ideation process for creative thinking and insight generation.

Below we present a step-by-step tutorial on using the Insight Recommender system, you can also read our post on the network ideation methodology, which offers a more in-depth discussion of the subject.

Video Introduction and Tutorial:

Workflow Outline:

This is a general outline of generating insight from your ideas. You can use it with InfraNodus, but we present it in general terms, so that you can also use it without any software at all as a methodology.

1. Create a new empty text network graph

2. Start writing or copy and paste an existing text draft 

3. Open the AI module and read a recommendation it provides, e.g.

    a) "explore a connection between these ideas / topics"
    b) "explore a certain topic deeper"
   usually, it will recommend you to make new connections between your ideas at the initial stage of your discourse evolution
4. Link the topical clusters 
    As you write, your ideas will separate into topical clusters (seen in the Analytics panel). Try to link them together with new ideas. If it's too hard, use the built-in GPT-3 AI to generate some ideas that may help you come up with your own.
5. Hide the most obvious concepts to see the context around them
   Sometimes the words and concepts you tend to use a lot will obfuscate the nuance around them. Select those ideas to temporarily hide them from the graph to let other, less represented ideas, emerge.
6. Bridge the structural gaps
   Find the gaps in the structure of your thinking and try to think of the connections between those gaps.
7. Disrupt / connect
   If your ideas are too interconnected, try to break them apart and generate some gaps. If they are too dispersed, try to connect them
8. Reiterate
   Oscillate between the different approaches above, so you can easily  shift between zooming in / zooming out on your ideas, connecting them and disrupting them at those different scales. 

Step-by-Step Insight Generation Workflow

Step 1: A Clear Page (Graph)

Open a new graph, a helper will appear suggesting you to add information into this graph. You can type it into the editor or import external data (files, spreadsheets, Google, Twitter, Evernote, RSS feeds, etc)

Step 2: Start Writing

Start typing in a text, or add an already existing one. If you add only one statement, our recommender system will propose you to think further of an interesting connection you could make to develop your ideas further. This is done because the existing structure is relatively homogeneous (mindviral immunity: low, discourse structure: biased — point 1 in the scheme above), so InfraNodus suggests to make it a bit more diverse this way:

Note, that you can always click the "Proceed" button if you want InfraNodus to demonstrate what needs to be done and to skip to the next recommendation step.

Step 3: Make a New Connection 

Following the suggestion, we click the "Proceed" button and InfraNodus highlights a possible connection, encouraging us to add a new statement that would further develop the two different topics it identified. After we do that, it will suggest that we look at the Analytics panel and the graph, highlighting the gap between the two statements that we added:

Step 4: Link the Topical Clusters

We can then add a new idea that would link those two clusters, making the discourse more connected. When you want to get further recommendations, click the "Proceed" button:

In this particular step InfraNodus recommends us to think of a possible connection between the structural gap it identified in our discourse. Bridging structural gaps in a discourse helps generate new ideas, so this particular recommendation may bring in a new idea and make the discourse more coherent.

Step 5: Reveal the Non-Obvious

After a few iterations the recommender system will propose to you that you delete the most influential nodes to see what's hiding behind them. This is a very interesting approach that allows you to develop the less represented parts of the discourse, giving attention to the parts of your text that could be abandoned otherwise:

Step 6: Bridge the Structural Gaps

The result, after we click "Proceed" is that the different topical groups are emphasised now, providing us more ideas to develop the ideas further. Note, that once you do that, InfraNodus may propose you to identify the structural gap between the two clusters of nodes that could be connected but are not yet (what you normally find in the Analytics > Insight panel) leading to generate a new connection between your ideas that you haven't thought of before.

Step 7: Disrupt / Connect

This is, probably, the most important stage of the process. At some point InfraNodus will detect the structure of your discourse and if it is too connected, it will propose you to disrupt the connections. If it's too disrupted, it will propose you to make the connections between the different islands of meaning that you create. In that it follows the model shown on Figure 1 above, where it takes you on the journey along the singularity / multiplicity / disruption / connection matrix.

Step 7: Reiterate

Continue and reiterate, following InfraNodus suggestions, until you feel like your discourse covers all the most important topics. Once you finish the process of free-writing using the graph, you can click on the "Interpret" panel and create a coherent text from your notes and ideas.