The hardest constraint in this assignment is the fixed-option format: without free text, the options themselves must carry the psychological weight. A question is only as good as its options — if the options feel identical or forced, the employee clicks through and learns nothing.
Rotter's (1954) locus of control is not a binary — it's a tendency that manifests in how people explain events, especially adverse ones. Dweck's (2006) growth mindset work adds the layer of whether ability and outcomes are seen as fixed or shapeable.
The key insight I built around: locus reveals itself in attribution, not in outcome. A good day can still be attributed externally ("I got lucky"). A bad day can still be owned internally ("I chose how I responded"). So the opening question routes by valence of day (word choice: Productive/Mixed vs Tough/Frustrating), but the follow-up question tests attribution — making the word matter less than the answer.
The options were stress-tested against this question: "Would someone who had a genuinely hard day, and genuinely felt out of control, feel like option B is describing them accurately — not as a judgment, but as a mirror?" I revised until the external options felt honest rather than shameful.
Campbell et al.'s (2004) work on psychological entitlement is useful here because entitlement is invisible to the person holding it. Nobody thinks "I am entitled." They think "I deserve this, and it's fair." So the questions can't name the thing directly.
Organ's (1988) Organizational Citizenship Behavior literature gave me the key construct: discretionary effort — doing things that aren't required. The contribution questions anchor on this. The entitlement questions anchor on the gap between effort and recognition — because that gap is where entitlement most commonly surfaces.
The follow-up questions (A2_Q_CONTRIB and A2_Q_ENTIT) are intentionally harder. A2_Q_CONTRIB asks the contributing employee to look underneath their generosity — was any of it transactional? A2_Q_ENTIT asks the employee who felt undervalued whether they moved on or whether the wound persisted.
Maslow's (1969) self-transcendence is the least-known of his hierarchy extensions. The core idea is that psychological health at its peak involves shifting from "what do I need?" to "what does the world need from me?" Batson's (2011) perspective-taking research adds the mechanism: the capacity to imagine another's inner experience — not just their situation.
The opening question (who comes to mind) is deliberately about who, not what. This catches something the employee can't manufacture: their spontaneous mental frame. Option A ("mainly me") isn't wrong — it's just narrow. Options B, C, D widen progressively. The follow-up questions then deepen whichever frame the employee signaled.
The assignment's sample tree used a single-layer branch. I designed two layers per axis: the opening question routes to one of two substantive question tracks (high/low valence for Axis 1, contribution/entitlement for Axis 2, self/altro for Axis 3), and then each track accumulates signals before a decision node selects the reflection.
This means someone who enters on the "good day" track but attributes their success externally still reaches the external reflection. The path adapts to the psychological signal, not just the surface framing of the day. This cost more nodes (9 decision nodes instead of 4) but made the branching more honest.
Trade-off I accepted: With 2 questions per axis and 4 options per question, a single answered question determines the signal. There's no signal averaging across questions. A richer version would weight signals across both questions and require a clearer majority — but this added complexity I couldn't justify for the current scope.
Both tracks in each axis converge to the same pair of reflection nodes (A1_R_INT / A1_R_EXT). This means the reflection text is axis-level, not track-level — a good day employee and a hard day employee who both land on "external" locus see the same reflection. This is a deliberate choice: the reflection should speak to the axis conclusion, not the surface path.
Rather than a single generic summary, I designed 8 variants keyed to the 2×2×2 combination of axis dominants. This keeps the summary in the data file (not hardcoded in the agent) and ensures every combination gets a specific, relevant close. The agent looks up the key axis1.dominant|axis2.dominant|axis3.dominant at runtime.
More axis-crossing questions. The current design treats the axes as sequential but largely independent. A richer tree would cross-reference: "You said you felt out of control (Axis 1 = external) — did that affect how generous you were able to be today (Axis 2)?" This is the progression the assignment hints at: each axis insight should seed the next question.
Richer signal accumulation. Currently, signals from A1_Q_HIGH and A1_Q_LOW overlap at A1_D2/A1_D3. I'd extend this to accumulate 3-4 signals per axis across both questions, giving the dominant calculation more data and reducing the weight of a single answer.
A "surprise" branch. One design I considered but cut for scope: a branch triggered when the axis-1 and axis-3 signals conflict (e.g., strong internal locus but self-centric radius). The insight here — from Maslow — is that agency without other-orientation produces effectiveness without meaning. A tree node that names that specific tension would be powerful.
A "return visitor" note. The current tree has no memory across sessions. A future version would compare today's dominant signals to yesterday's and surface the delta — "Yesterday you leaned external. Today, internal. What changed?"
- Rotter, J.B. (1954). Social learning and clinical psychology. Prentice-Hall.
- Dweck, C. (2006). Mindset: The new psychology of success. Random House.
- Campbell, W.K., Bonacci, A.M., Shelton, J., Exline, J.J., & Bushman, B.J. (2004). Psychological entitlement: Interpersonal consequences and validation of a new self-report measure. Journal of Personality Assessment, 83(1), 29–45.
- Organ, D.W. (1988). Organizational citizenship behavior: The good soldier syndrome. Lexington Books.
- Maslow, A.H. (1969). The farther reaches of human nature. Journal of Transpersonal Psychology, 1(1), 1–9.
- Batson, C.D. (2011). Altruism in humans. Oxford University Press.