Technology probes yield countable logs; design probes yield works that need interpretation

What it is

A technology probe’s main yield is countable logs: who did which action when, for how long, whether it failed. A design probe’s main yield is work that needs subjective reading: photographs, maps, handwriting, collage. Neither writes requirements directly, but they enter analysis on wholly different paths. Scoring postcards as questionnaire items, or reading logs as “the user’s voice,” sends the wrong downstream method at the output type. Logs answer what happened and how often; works answer how it was felt and figured.

Why it happens

Logs are written by a device under rules. Units exist in advance (event, session, error code), so they can be summed, plotted, and confronted with interviews. They do not contain meaning: a late-night tap may be a slip, anxiety, or a night shift. Works are chosen by a person in a medium and a moment; meaning is dense and not summable—two photographs of “hating the alarm” have no mean. Reading them, the designer’s priors, visual literacy, and project agenda help constitute what was “seen.” So the technology-probe risk is taking counts for causes; the design-probe risk is taking being moved for evidential strength. Analysis has to follow the yield: structure then meaning, or meaning then follow-up—not one codebook for both.

Studying it

For technology probes, define an event dictionary and missingness rules in advance; describe and flag anomalous sessions before sampling for contextual interviews. For design probes, do close readings and designer memos first, then see whether contestable themes appear; themes are after the fact, not a scale. Mixed deployments should align log timestamps with work dates for follow-up, not for mutual proof. Gaver refused to fold cultural probes into frequencies; Hutchinson and colleagues treated logs as a quantifiable use layer. In review, keep two piles; do not put “12 times a day” and a poetic photograph on one slide as one finding.

Where it stops holding

Logs are not always cleanly countable: clock drift, shared accounts, and children mashing destroy the unit. Works are not always uncountable: if the task is “one a day,” counts still describe participation, though content still wants reading. Voice memos left on a technology probe sit closer to work. Structured diary scales slipped into a design probe should be analyzed as survey, not as probe. Legal forensics need auditable logs; design-probe work usually will not qualify.

Applying it

• At kickoff, name whether the main yield is logs or work, and name the matching analysis owner.
• From logs, produce a timeline and anomalies before deciding what to ask; from work, produce reading memos before deciding whether to count participation.
• Do not score work on Likert items, and do not use log rows as quotations unless there is contemporaneous audio.
• Deliver in two volumes: counts and readings, cross-referenced, with conclusion levels kept apart.

Related

• Same group: Q5.08.1 Technology probes typically last weeks, shorter than the incubation of design probes · Q5.08.3 The material quality of a probe pack shapes engagement and return rate · Q5.08.4 Probes suit early inspiration, not validating a specific design
• Adjacent: Q5.05 Probes · Q4.02 Coding and thematic analysis
• Search terms: probe output types · instrumented logs · interpretive returns