InfraNodus algorithm promotes cognitive variability. In simple terms, if your discourse is too biased, it will steer it toward diversification. If it's too dispersed, it will steer it towards more focus.
Practically, this is done by promoting variability in your thinking along the 4 axes.
1. Focus <-> Exploration
and
2) Zooming out <-> Zooming out
You probably know how useful it is to shift from one modality to another in your own research and thinking process.
When we are writing a new paper, for example, we want to be able to zoom in on a topic, but we also want to have access to the whole picture (zooming out), in order to place our findings in the current context and offer something innovative and useful to the others. At the same time, we also need to shift between moments of focus and exploration to allow for the new ideas to come in, but to also produce something concrete from time to time.
InfraNodus promotes this way of thinking by analyzing the structure of your thought. It will then detect how long you stayed in one state (e.g. zooming in and focused) and propose a change (e.g. zoom out or explore new ideas), to guide your thinking process.
We call it ecological thinking, because it encompasses different stages: from biased to dispersed.
Here is how it works technically.
Variability of Thinking and Text Network Structure
Using a combination of text network modularity (i.e. how pronounced the community structure is in the graph), distribution of influence (how well the most influential concepts are spread across the topics), and narrative dynamics (influence propagation and degree distribution), InfraNodus identifies the current state of the discourse structure. It will then provide AI-powered recommendations to steer this discourse towards more adaptive dynamics, which can help develop an idea, if needed, or get a good overview of a subject.
Here is how it works, you can also use the mandala as a reference image.
There are four main states that the dynamic structure can be in:
1. Biased (Uniform) [stages 8 → 1 → 2]
2. Focused (Regular) [stages 2 → 3 → 4]
3. Diversified (Fractal) [stages 4 → 5 → 6]
4. Dispersed (Complex) [stages 6 → 7 → 8]
InfraNodus' AI-enhanced recommendation algorithm will guide the user through these different states of the discourse structure, giving more weight to the "diversified" (fractal) state.
Each of the states leads to the emergence of certain qualities that may be useful depending on the context and on the objectives. For instance, a biased state may be useful for large-scale persuasion, propaganda, or publicity. A focused state may be useful for an informative news article. A diversified state may be useful for an extended report or a scientific discourse (so that it stays open and objective). A dispersed state is particularly useful for brainstorming sessions, fiction, and poetry.
Below we describe some of the qualities of each state. While InfraNodus by default will steer the discourse towards the diversified (fractal) state, you can also set it to have a different set of priorities, depending on your objectives.
8 → 1 → 2 State: Uniform Dynamics (Bias)
Characterized by uniform dynamics, which is usually operating on small- to mid-scale, focusing on one topic.
The discourse will either have no distinct community structure. Rather, a central idea/node to which everything else is connected. Or unequal distribution of influence within a discourse (so that more preference is given to a specific idea to which everything else is subjugated). Rhythmical variability.
Pros: This state may be useful for building an idea from scratch, constructing something new, as a rhetorical device, for propaganda or publicity.
Cons: Missing the whole picture, too much focus on local detail.
8→1 stage: #bias #genesis
Finding a starting point within a multiplicity of ideas, zooming into a dispersed entity and starting to focus on specific ideas and connecting some of them together.
1→2 stage: #bias #growth #exponent
Here we operate at the smallest scale, starting from a single concept or a point. As it's the only idea that exists, this topic is biased. We start to link other ideas to it (a mix of exploration 80% and focus 20%), increasing the focus, decreasing the exploration. We also shift towards a bigger scale, zooming out (80% zoom in, 20% zoom out), taking in a bigger range of ideas and linking them together. Here is where our initial ideas grow and take shape. It is easy to stay biased at this point, so in order to avoid that one has to occasionally bring in external concepts and also make jumps between different ideas that are forming.
2 → 3 → 4 State: Regular Dynamics (Focus)
Characterized by regular dynamics. There is a focus on a certain idea, but there is also some diversity on the global level. Zooming out and taking into account the global objectives while also maintaining the main focus. The narrative dynamics is more or less predictable. We can easily understand what the discourse is about and how it may evolve.
Pros: This state may be useful for communicating a clear idea and building a case or a focused narrative.
Cons: Reaches saturation quickly, limited resources for developing an idea further. Rigid, not adaptive.
2→3 stage: #focus #saturation #plateau
Stage 2 is perfectly fractal. We connect the ideas we already had and bring in new ones. We are at the 50/50% mix of small scale and big scale, exploration and focus. This means we can look at specifics, but also get an overview. Develop a very specific topic, but also connect different topics together. This will bring the discourse towards a focused state (stage 2 to 3). Higher connectedness, focus on multiplicities, increasing the scale, zooming out, linking ideas more than exploring (80% to 20%).
3→4 stage: #focus #conservation #intensification
We try to connect all the ideas we have. We also start zooming in again, decreasing the scale, connecting ideas within a smaller area, deepening the focus inside the existing structure (stage 3 to 4 on the graph).
4 → 5 → 6 State: Fractal Dynamics (Diversified)
Characterized by fractal variability: the narrative qualities will alternate in a self-similar way on multiple scales. In simple terms, it shifts the subject both on the local and on the global levels, there are moments of exploration but also of focus.
Pros: Adaptive and diversified state, useful for communicating an objective perspective that can assimilate multiple points of view.
Cons: As this is more of a self-reflective state, it is less suitable for communicating a clear agenda across. Lacks singularity (if it is needed).
4→5 stage: #diversification #assimilation #release
Once the discourse has become even more "focused" (stage 4 to 5), InfraNodus will also suggest to develop the specificities, zoom in further, go bigger scale, and, thus, reduce (global) focus and increase the proportion of exploration again (to 20%).
5→6 stage: #diversification #redirection #fractalization
We continue bringing in more new ideas on the local level and also on the global one as well. Making new connections that didn't exist before. Bridging the structural gaps in the network. InfraNodus will propose to diversify the discourse by zooming out and exploring more clusters, developing each of them further (stage 5-6: moving between locally related ideas and also jumping across the structural gaps in the graph).
6 → 7 → 8 State: Complex Dynamics (Dispersed)
This stage is characterized with large-scale, dispersed, exploratory thinking. The best analogy is a dream state. The connections are rewired in new ways, ideas are reorganized, the old ones are expelled, the new ones come in. There are multiple structural gaps forming, which can offer opportunities for new ideas to emerge.
Pros: Useful for finding new perspectives and innovative ideas. Poetic.
Cons: May seem too dispersed and lack focus.
6→7 stage: #dispersion #reogranization #dream
At some point, we shift towards dispersion by giving more weight to the exploration process, rather than focus (state 6-7: dispersion). Breaking the links, disrupting connections, finding and generating more gaps, on a bigger scale.
7→8 stage: #dispersion #reset #
Once the scale is large enough, we zoom into the small scale again while exploring until we find a new concept we'd like to develop (stage 7-8).
8→1 stage: #bias #genesis
We then focus on it and develop it further again (stages 8 to 1). Reiterate.
Video Tutorial:
Detecting Bias in LLM Output:
Comments
0 comments
Please sign in to leave a comment.