Back to Home

Master's Thesis in Digital Design · DD202A

AI‑Assisted Overviews for Family Coordination

An agentive, glanceable display using AI-generated portraits and calm technology principles

Kristianstad University (HKR) · 2026

Prototype screenshot — replace with actual image

The Problem

Fragmented channels, not missing information

Parents of school-age children don't lack information about their week — they struggle to assemble it into a coherent picture. Coordination data is scattered across shared calendars, school communication platforms, messaging apps, and location-sharing services. Notifications arrive continuously, often detached from the moment they're relevant. The result is not an information deficit but a retrieval and assembly problem that demands significant cognitive effort.

Fragmentation & Retrieval Effort

Information relevant to coordination was distributed across school apps, messaging channels, email threads, and shared calendars. Parents described searching for the same piece of information multiple times, or relying on memory because finding the source felt harder than guessing.

Theme from 6 semi-structured interviews

Overview Without Depth

Parents wanted “one place” to understand the week — not to centralise all details, but to reduce the cognitive effort of consulting many sources. The goal was a quick sense of who was doing what, when, and whether anything needed attention.

Theme from 6 semi-structured interviews

Coarse-Grained Cues

Phrases like “on the way home,” “still at school,” or “free later” were described as more useful than exact locations or timestamps. What mattered for coordination was whether a plan was feasible or someone was available.

Theme from 6 semi-structured interviews

Predictable, Bundled Updates

Parents valued low-frequency, predictable updates — morning summaries, pre-evening check-ins — over constant micro-notifications. Older, slower forms of communication were referenced positively: paper schedules on the fridge, a shared whiteboard in the hallway.

Theme from 6 semi-structured interviews

The Concept

An agentive, glanceable overview

An ambient display that represents each family member through AI-generated portraits and contextual attributes. Attributes — clothing, objects, accessories, environmental cues — serve as visual shorthand for the week's activities. When an event needs attention, it surfaces through the portrait. When someone claims responsibility, it visually resolves. The display sits in a shared domestic space: a wall, a shelf, a counter where family members pass.

Prototype screens showing portrait states — replace with actual image

Portrait + Attribute

Identity is separated from context. Each family member has a stable portrait that serves as a visual anchor. Layered onto this are variable attributes: a tennis racket signals sports practice, a school desk signals exams, a packed bag signals travel. These visual shorthands operate at the level of feasibility judgment, not scheduling precision.

Temporal Layering

The display creates visual depth across the week. Immediate activities appear prominently in the foreground. Approaching events emerge gradually in the background. As the week progresses, completed events recede and upcoming ones move forward — a soft transition model that avoids the discrete state changes of notification-based systems.

Claimed = Handled

A family member doesn't need to know how something is handled — only that it is. When someone claims an event, it visually resolves in their portrait. A binary model (needs attention vs. claimed) replaces multi-level urgency, reflecting the interview finding that coarse-grained cues are sufficient for coordination.

The Role of AI

AI operates as infrastructure, not a visible agent. It aggregates data from distributed sources, categorises events into the family's visual vocabulary, determines temporal weighting, and generates portrait variants. What it does not do: suggest schedule changes, resolve conflicts, infer intent, or send notifications.

Design Process

Concept-Driven Interaction Design Research

The study follows the CDIDR framework (Wiberg & Stolterman, 2010), where a theoretical concept — rather than an identified user need — serves as the starting point for design inquiry. The guiding concept was articulated from the convergence of three research streams: family coordination as socio-material practice, calm technology and peripheral awareness, and agentive AI systems.

Six qualitative interviews with parents of school-age children, supported by sketch-based probes, established the empirical foundation. This fed into iterative concept development, internal critique documented through a design log, and external critique through design supervision and a technology probe deployed with two families over two days.

Sketches, iterations, and pivot documentation — replace with actual images

Key Insights

What the design process revealed

From Notification Filtering to Weekly Overview

The original framing centred on interruption management. Interviews revealed that families had already learned to cope with notification volume. The real need was not filtering interruptions but assembling a coherent sense of the week from scattered sources.

Group Portrait to Individual Portraits

Three approaches to composite family portraits via API were attempted and abandoned. Single-image generation lacked compositional control. Separate renders introduced inconsistent scale. The pivot to individual portraits opened finer-grained per-person state communication — a design finding in itself.

Conversational vs. Programmatic AI

Conversational prompting (ChatGPT-style) achieved compositional nuance that API-driven generation could not reproduce. The gap between what AI can do conversationally and what it can produce programmatically is itself a contribution of the design process.

Colour Overlay as Coordination Cue

Icon or emoji overlays broke the calm impression — they felt like notifications imposed on the image. A reddish colour overlay proved more coherent: the portrait itself shifts tone rather than having an alert stacked on top, staying true to the ambient quality the display aims to achieve.

Design Principles

Three principles from theory and practice

Design for judgment, not precision

Parents assess feasibility rather than exactness. The prototype's use of coarse-grained attributes over detailed schedules reflects this, aligning with calm technology's emphasis on abstraction over exhaustive detail.

Use AI to reshape information, not to control decisions

Participants valued assistance that clarified rather than automated. The prototype positions AI as a mediating layer that aggregates and re-presents without inferring intent or making scheduling decisions — consistent with human-centred AI frameworks.

Support shared rhythms through ambient placement

The preference for predictable, bundled updates and the positive response to ambient placement suggest that coordination tools benefit from operating on domestic temporal rhythms rather than stochastic notification regimes.

Design Space

Five dimensions for AI-assisted overviews

Abstraction of StatusTemporal ScopeDegree of AgentivenessModality & PlacementInteraction Posture

The four interview themes were translated into five design dimensions that map how an AI-assisted overview differs from current app-based coordination tools. A fifth dimension — degree of AI agency — emerged from the overall pattern: parents wanted assistance that clarified rather than controlled. Together, these dimensions frame how an overview can remain peripheral, contextual, and supportive of shared awareness.

Theoretical Grounding

Three converging research streams

Calm TechnologyFamily Coordination (CSCW)Agentive SystemsPeripheral AwarenessResearch through Design

Family coordination research shows that domestic scheduling is a socio-material practice sustained through negotiation, environmental cues, and peripheral awareness — not through centralised information systems. Calm technology offers an alternative to interruptive delivery by allowing information to move between centre and periphery as needed. Agentive systems research explores how AI can mediate attention without replacing human judgment.

These three positions share a common orientation: human judgment over system automation, ambient availability over interruptive delivery, and sufficient abstraction over exhaustive detail. Together they define the theoretical space this project inhabits.

Research Questions

What this thesis investigates

RQ1

How can an agentive overview of family commitments be designed to reduce perceived channel fragmentation and notification overload for parents of school-age children?

RQ2

In what ways might AI summarise and re-present distributed family schedule information to support quick, shared decisions without requiring focused engagement?

Fredrik Sjöberg · HKR 2026

Supervisor: Kristoffer Åberg

Design supervision: Mattias Rylander

Examiner: Martin Wetterstrand

Master's Programme in Digital Design (DD202A), Kristianstad University