Using AI in subjective data mining
As a designer interested in Data Viz & AI, I set out to catalog & visualize my journals, using Soft Formulas created with ChatGPT to extract subjective data from my journals.
How does this work?
Using these subjective formulas, I was able to pull data from my journals in a replicable process, which is something that would’ve been entirely based on gut instinct beforehand.
-
This measures the amount of hope that is shown in any given text.
H=(P x O) + (R x B) - (D ÷ A)
H: level of hope
P: Perceived Possibility (1-10)
O: Optimism (1-10)
R: Resilience (1-10)
B: Belief in Change (1-10)
D: Doubt (1-10)
A: Awareness (1-10)
-
This measures the intensity of the experiences.
E = (M x V) + (R x T) - (S ÷ A)
E: Emotional Intensity
M: Magnitude (1-10)
V: Valence
R: Resonance (1-10)
T: Time (in days)
S: Self-regulation (1-10)
A: Awareness (1-10)
-
This measures the balance of longing and contentment throughout my writings.
LC = (D x Y) + (W x T) - (G ÷ A)
LC: Longing or contentment score
D: Desire Factor (1-10)
Y: Yearning (1-10)
W: Words expressing peace (1-10)
T: Trust in the Process (1-10)
G: Gratitude (1-10)
A: Awareness (1-10)
-
Or FSI for short. This measures the text I wrote that can be tied back to my faith as a Christian.
FSI = (Fw x 2) + (Sp x 1.5) + (Pt x 1) + (Ec x 2.5) all divided by total amount of words in the entry
F = Faith-related words
S = Scripture references
P = Personal reflections
E = Emotional context
Using Dall-E and my Subjective Formulas, I explored multiple data visualization concepts (topographic Maps, Soundwaves, constellations, Mesh Clouds, and Sankey Diagrams).
The below renderings result from my “training” Dall-E on the data I gathered using my subjective formulas. These renderings are abstractions of 2D graphs. The peaks and valleys show the emotional intensities, the colors show the tone of the memory, and lower reflective surfaces symbolize how we reflect on the past and often remember it differently. Additionally, the ribbons are made of text from my journals, allowing viewers to be immersed in the data in a poetic way and interact with it.
The data
Before the DALL-E renderings, there were endless charts of data that the bot was able to pull.
What next?
This exploration is part of an ongoing process of seeing how STEM Laws & Formulas can be reimagined for subjective data. Many STEM laws describe objective phenomena, but when reinterpreted, they can model subjective experiences like emotions, memories, and personal perceptions. Click the collapsable menu options to see what I’m currently exploring.
-
Original Use: Break down signals like sound or light waves into their frequencies.
Subjective Twist: Treat emotions, thoughts, or memories as waveforms and analyze their dominant "frequencies." -
Original Use: Measures disorder in a system—entropy always increases over time.
Subjective Twist: What if we measured the entropy of thoughts, memories, or emotions? -
Original Use: Describes how quantum states evolve, where particles exist in multiple states until measured.
Subjective Twist: Represent complex, conflicting emotions as "quantum superpositions." -
Original Use: Describes the path of least resistance in physics.
Subjective Twist: Model how the mind reconstructs memories by choosing the "easiest" path through past experiences.
Importance
What about this process makes it more than just a fun experiment?
It bridges A.I. + emotion + humanness
Bringing this humanness into AI left me with many surprises. What surprised me the most was how my emotions connected; hope often followed uncertainty, while reflection appeared most after periods of struggle.
Instead of just stating that a period of my life was “hopeful” or “introspective,” my formulas quantify the shifts in emotion, faith, and perception over time.
It highlights trends
Because my journal entries span an entire year, these formulas highlight trends I may not have even noticed myself. Most A.I. and data science approaches struggle to work with deeply personal, introspective text because:
They focus on hard metrics (word count, sentiment polarity, frequency analysis).
They struggle with emotional nuance—how faith is expressed differently in struggle vs. celebration.
My project is a fusion of human interpretation + A.I. assistance, making these subjective concepts repeatable & scalable.
Instead of just “word frequency” (e.g., how often I say “faith” or “hope”), my formulas account for context—whether faith is expressed as doubt, surrender, or confidence.
It benefits design higher education
With data visualization, the challenge is that traditional research focuses on objective, numerical data, but emotional, subjective data is just as valuable.
With my process, we can focus on teaching students how to quantify and visualize subjective experiences using custom formulas (e.g., Hope Index, Emotional Intensity Score, etc). Imagine a semester where students were able to create their own “mathematical” formulas & the learnings they could gain from this process.
It prepares them for an ever-changing field; not only can these skills help in existing fields such as UX/UI research or finding customer insights, but having even the smallest bit of fluency with these generative programs is important with how the design field is changing and may continue to change,
Why should we even bother with AI if it’s just a shortcut?
Ah, you’ve stumbled on a philosophical issue here: if someone’s first thought is: “How can I take shortcuts?” that is a problem with the person and not a problem with the tool. As designers, AI should be our Copilot, or a sparring partner of sorts. Nothing more. AI isn’t going away anytime soon, and I believe it is the responsibility of Design Academia to be engaged with emerging technologies. That’s not something that you can do from the sidelines, so you need to get your hands dirty.