Satisfaction to Altruistic Voting Concept

In the current model, dissatisfaction does not only describe how well agents fit the environment. It also shifts how likely they are to cast altruistic rather than self-regarding ballots.

1. Behavioral Design Choice

Agents do not learn ballot strategies directly. The learned/updated behavioral axis for the thesis is participation.

Ballot mode is still heterogeneous per vote, but controlled through a state variable (altruism_factor) rather than strategic optimization.

2. Dissatisfaction Signal

Before election each step, each agent computes dissatisfaction_value as distance between:

• agent preferred distribution (personal_opt_dist)
• target distribution from satisfaction_mode

Supported target modes:

• area (default and used in thesis runs)
• global
• knowledge
• combination

Baseline trajectory:

• dissatisfaction baseline is EMA-updated
• dissatisfaction_signal = dissatisfaction_value - dissatisfaction_baseline

In thesis baseline, dissatisfaction level itself is used to map altruism in satisfaction mode. It is not learned nor strategic.

3. Satisfaction-Mode Mapping to Altruism

If altruism_mode=satisfaction, the mapping is:

• s = 1 - dissatisfaction_value
• target = sigmoid(k * (s - theta))
• with response smoothing parameter gamma

Then altruism_factor is clipped into configured bounds.

Higher satisfaction tends to raise altruism probability, while lower satisfaction tends to lower it.

4. Vote-Time Mode Selection

During vote casting, a Bernoulli draw is made using altruism_factor:

• altruistic mode with probability altruism_factor
• self-regarding mode otherwise

Ballot construction:

• altruistic mode: use estimated distribution from limited knowledge (known_cells), convert to ordering, then score options against that ordering
• self-regarding mode: use static precomputed oppose-scores from preference-group ordering

This keeps vote-mode heterogeneity without turning the model into a strategic policy game.

5. Why This Concept Exists

The thesis needs a mechanism linking environment fit to collective-choice quality capacity. This mapping provides that channel:

• state fit (satisfaction) influences ballot orientation
• ballot orientation influences quality-gate success probability
• quality sign influences reward regime and subsequent participation learning
• fixed shift of satisfied voters to altruistic vote can limit lock-in states

6. Limits

This is not a model of explicit moral reasoning. "Altruistic" here means voting according to an estimated common-quality signal rather than self-regarding preference.