The ABCD in Multimodal Intrapersonal and Interpersonal Dynamics (MIND)

What is ABCD?

ABCD is a theory-grounded annotation framework used within the Multimodal Intrapersonal and Interpersonal Dynamics (MIND) taxonomy. It captures four core psychological dimensions that clinicians attend to when reading a patient's written or spoken language:

Each dimension is further split along two axes:

• Intra- vs Interpersonal: B and C each have a Self variant (B-S, C-S) and an Other variant (B-O, C-O), reflecting whether the psychological content is directed inward or outward.
• Adaptive vs Maladaptive: every sub-category is coded as either adaptive (healthy, functional) or maladaptive (distressed, dysfunctional).

The full schema therefore yields six annotation dimensions: A, B-O, B-S, C-O, C-S, D.

A: Affect

Affect captures the type of emotion expressed by the writer. It is the most granular dimension, with 14 distinct sub-categories.

Adaptive affect

Note on A-3: Sadness and grief are coded as adaptive when they are proportionate to circumstances (e.g. mourning a loss), rather than pervasive and disconnected from any trigger.

Maladaptive affect

The odd/even numbering is intentional: odd codes are adaptive, even codes are maladaptive, making the pairing visually explicit in annotations.

B: Behavior

Behavior captures what the writer does: actions rather than feelings or thoughts. It splits into two sub-dimensions depending on the target of the action.

B-O: Behavior toward Others

Relating behavior (B-O-1) encompasses warmth, cooperation, and healthy connection. Its maladaptive mirror, B-O-2, covers aggression, withdrawal, or avoidance when facing interpersonal stress.

Autonomous control behavior (B-O-3) reflects healthy boundary-setting and self-direction in relationships; B-O-4 flips this into rigidity or coercive control.

B-S: Behavior toward Self

This dimension collapses to a single adaptive/maladaptive pair because the relevant behavioral space is narrower when the actor and target are the same person: either the writer is taking care of themselves, or they are not.

C: Cognition

Cognition captures how the writer perceives their world: the mental representations they hold rather than the emotions they feel or the actions they take. Like Behavior, it splits into an Other-directed and a Self-directed variant.

C-O: Cognition toward Others

C-O mirrors the structure of B-O: both dimensions are organised around the twin psychological needs of relatedness and autonomy, a pattern that will resurface in D.

C-S: Cognition toward Self

A writer high in C-S-2 might describe themselves as worthless, fundamentally broken, or undeserving of care: internal narratives that clinical approaches such as CFT and CBT directly target.

D: Desire

Desire captures the writer's underlying motivation: the need, expectation, intention, or fear that drives the psychological content in a passage.

Adaptive desires (needs the writer expects to meet)

These three map cleanly onto Self-Determination Theory's three basic psychological needs: relatedness, autonomy, and competence.

Maladaptive desires (needs the writer fears will not be met)

The maladaptive Desire codes are not simply "wanting bad things": they are thwarted versions of the same healthy needs. A writer coded D-2 still wants connection; they simply do not believe they will ever have it.

The Full ABCD Picture

Putting it all together, a single annotated passage might receive codes across multiple dimensions simultaneously:

"I don't know why I even try. Nobody cares about me anyway, and I can't seem to do anything right."

A possible annotation:

The richness of this multi-label representation is what makes ABCD well-suited to NLP tasks that go beyond binary distress detection: it supports fine-grained triage, progress tracking, and interpretable model outputs.

Why ABCD Matters for Computational Work

Classical mental-health NLP relies heavily on keyword lists and binary sentiment. ABCD offers three advantages:

1. Theory grounding: every label traces back to established clinical constructs (SDT, attachment theory, CBT/CFT), making the taxonomy interpretable to practitioners.
2. Adaptive/maladaptive distinction: the same surface emotion (e.g. anger) can be healthy (A-9) or harmful (A-10) depending on context; models trained on ABCD must capture this nuance.
3. Multi-dimensionality: a passage can simultaneously carry affect, behavioral, cognitive, and motivational signals, enabling richer downstream tasks such as treatment matching or longitudinal change detection.

Future posts will show how we operationalise ABCD labels in the CLPsych shared task and what NLP architectures best recover each dimension.

Reference

Atzil-Slonim, D. (2025). Multimodal Intrapersonal and Interpersonal Dynamics (MIND): A Transtheoretical Coding Manual. https://doi.org/10.17605/OSF.IO/SJE8C

Visualize the ABCD

The interactive plot below shows keyword clusters for each ABCD sub-category across Twitter-BERT model. Use the sidebar to filter sub-categories or colouring modes.