kureshik.fun

Method

The Scientific Interest Index (SII) quantifies how intriguing or research-worthy an object, concept, or phenomenon is. It combines several normalized features — Novelty, Salience, Explanatory Potential, Impact, and Tractability — into a single score between 0 and 1.

Each feature is evaluated on a scale [0, 1] based on semantic, contextual, or statistical analysis.

Features

Symbol	Name	Meaning	Range
N	Novelty	How rare or unexpected the object is	[0, 1]
S	Salience	How much the object stands out or attracts attention	[0, 1]
E	Explanatory Potential	How strongly the object can generate or expand scientific hypotheses	[0, 1]
Imp	Impact	How significant or influential the object can be across disciplines	[0, 1]
T	Tractability	How feasible it is to study or experiment with the object	[0, 1]

Formula

1. Weighted Linear Model

\text{SII} = ( w_1 N + w_2 S + w_3 E + w_4 Imp ) \times T^{\alpha}

$w_i$ : feature weights, $\sum w_i = 1$
$\alpha$ : parameter controlling the influence of tractability

2. Multiplicative Model (optional, more selective)

\text{SII} = N^{a} \cdot S^{b} \cdot E^{c} \cdot Imp^{d} \cdot T^{\alpha}

where exponents $a, b, c, d, \alpha$ define sensitivity to each factor.

Training the Weights

Given a dataset of objects with expert-rated interest scores $y_i \in [0,1]$ :

\hat{y_i} = \sigma(w_0 + w_1 N_i + w_2 S_i + w_3 E_i + w_4 Imp_i + w_5 T_i)

Weights are optimized by minimizing Mean Squared Error (MSE):

L = \frac{1}{n} \sum_{i=1}^n (\hat{y_i} - y_i)^2

Example

Object	N	S	E	Imp	T	SII
Stone	0.05	0.02	0.01	0.01	0.99	0.02
Elephant in a room	0.9	0.95	0.6	0.4	0.3	0.21