Abstract low-poly geometric background in shades of pale blue and white

Data in the Middle: The common language of research

What IASSIST 2017 taught us about documenting qualitative data

Between 23 and 26 May 2017, the University of Kansas in Lawrence hosted IASSIST 2017 under the theme “Data in the Middle: The Common Language of Research.” The archived gathering brought together work on big data, deep learning, digital agriculture and global food security, alongside the practical routines that make research programmes usable. Its central lesson for qualitative researchers was simple: data become shareable only when their context, decisions and meanings can be followed by someone who was not in the room.

For Australian audiences, that lesson carries particular force. A transcript from a regional health service, an interview with a dairy farmer in Gippsland, or observations from a First Nations community project may travel well beyond its original team, but it cannot travel safely without records of who spoke, how consent was framed, what the researcher noticed, and how later interpretations were made. Good documentation does not merely describe files. It preserves the chain from evidence to claim and helps organisations meet expectations around transparency, governance and accountable reuse.

Documentation choice Main strength Best use Common risk
Field notes Closeness and ambiguity Early analysis Context mistaken for evidence
Interview transcripts Verbatim detail Quotation and close reading Incomplete de-identification
Codebooks and analytic memos Traceable reasoning Review and reuse Labels without rationale
Mixed project records Integration across formats Mixed-method projects Versions fall out of sync

Documentation is a research decision

Qualitative documentation is often treated as clerical work: notes are tidied, interviews are transcribed, consent forms are filed and a code list is exported. Yet each step encodes a research decision. Transcription choices alter rhythm and emphasis; inclusion rules change the case; a code can move from a descriptive label into an analytical claim. Recording those decisions makes the method inspectable instead of presenting the final interpretation as though it appeared naturally in the field.

The IASSIST theme, Data in the Middle, helps explain why this matters. Whether the subject was a large agricultural dataset, a machine-learning model or food-security evidence, credible reuse depended on shared conventions and transparent pathways between people, systems and institutions. Qualitative data need the same treatment. Their irregularity is not a defect to conceal. The irregularity may carry the evidence, provided the record shows what was observed, what was said, how categories were derived and which uncertainties remain.

Australian projects add another layer because collaboration often crosses institutions, states and sectors. A university team may work with a state agency, a producer group, an Indigenous governing body and a technology supplier, each with different expectations about access. A documentation plan made at the start creates common ground. It also avoids the common failure of treating consent as permission for every later use rather than as a set of conditions that must accompany the data.

Big data still depends on human context

Big data and deep learning can look worlds away from field interviews, but their interpretive gaps are familiar. A training set needs a record of source conditions; a model needs decisions about missing values, labels and validation. When qualitative evidence informs such systems, interviews, site observations and open-ended surveys can reveal whether variables mean what engineers assume. They can expose categories that perform badly for particular communities and show where a seemingly precise score depends on a fuzzy social concept.

Digital agriculture makes this connection concrete. Sensor readings may describe soil moisture, while a farmer's account describes why a paddock was managed differently, why an application failed, or how seasonal work was organised. Global food-security analysis depends on those situated details as well as national statistics. The archived conference program shows the range of topics considered in Kansas; the continuing lesson is that technical scale does not remove the need for human interpretation. It makes interpretation more traceable.

Deep learning raises another documentation issue: traceability back to examples. A qualitative sample can document the kinds of language, imagery or behaviour a model encounters, including cases that challenge its categories. It cannot substitute for validation, nor should researchers claim that interviews prove a model's performance. Its value lies in showing the social world around the data, surfacing ambiguity and giving future analysts reasons to question a label. These records belong beside code, model cards and data sheets, not in a separate qualitative cupboard.

From field notes to durable records

The most durable qualitative record begins before the first interview. File-naming rules, participant codes, consent versions and a decision log establish a predictable structure. Later, transcripts or observation notes can be linked to the protocol, setting and interviewer that produced them. This is less about rigid uniformity than stable identifiers. If two sites use different local practices, the record should preserve those differences while using shared fields to show where they occur.

Raw material and interpretation should remain distinguishable. Cleaned transcripts may be easier to analyse, but they can quietly remove hesitations, repetitions or interviewer prompts that matter to the argument. An audit trail can retain the original audio, a minimally edited transcript, corrections and a dated account of each transformation. For sensitive material, the archive may hold only a secure master or a carefully controlled derivative, provided the access decision itself is documented. “De-identified” is a process description, not a guarantee that re-identification is impossible.

Version control matters just as much. Shared drives are convenient, yet filenames such as final_final2 offer no research history. A simple log can record who changed a code definition, why a segment was excluded and which version fed a publication or repository deposit. Automated tools help, but they cannot decide why an ethical, linguistic or contextual judgement was made. That rationale remains part of the evidence and should be reviewed with the same care as a numerical transformation.

Making context searchable and reusable

Documentation should make qualitative data findable without pretending it is fully standardised. Searchable metadata can include research question, setting, participant group, recruitment route, collection period, language, interview format, consent basis and analytical purpose. These fields support discovery and responsible access. They should not expose identifying details in public catalogues. Access conditions, retention periods and restrictions on quotation belong in metadata where searchers can understand the limits before requesting a file.

Context can be structured, layered and linked. A structured record captures recurring fields; a narrative project memo explains why the research team made consequential choices; a codebook defines categories and permitted uses. Links among these elements let a reader move from a theme to its supporting material and from that material to the study's scope. The aim is not to compress lived experience into a database row. It is to retain enough provenance for another researcher to understand how the evidence was produced and what it can bear.

Quality assurance should be proportionate and transparent. Checking transcripts against recordings, reviewing code definitions with local experts, documenting translation choices and testing repository access all improve reliability. Australian English terms such as “interview”, “farm”, “project” and “data” may appear in a record alongside place names and community-specific vocabulary. Preserve the participant's wording, then supply a separate analytical translation where needed. A tidy keyword list cannot capture every meaning, but a clear glossary and decision history can prevent quiet distortion.

Lessons for Australian research practice

These principles become more concrete across Australian settings. In Canberra, research often sits close to Commonwealth, state and territory agencies, so records must clarify which institution owns a dataset, which agreement permits reuse and how public-interest obligations affect disclosure. Around Melbourne, CSIRO, universities, consultancies and social research firms regularly form project teams. Their different commercial and scholarly timelines reward documentation that lets work move between proposal, fieldwork, analysis and client or community reporting without repeated clarification.

Place also changes what counts as adequate context. A project near Toowoomba may compare several farming systems and involve growers who describe equipment, rainfall and labour in familiar local terms. Research with Aboriginal and Torres Strait Islander communities requires more than a standard consent box: community authority, cultural protocols and agreed data governance must shape the record. In remote Northern Territory settings, unreliable connectivity can make cloud-only documentation risky, so offline capture and a delayed synchronisation plan may be necessary.

The Australian research market makes this practical rather than merely theoretical. Universities, government agencies, CSIRO, industry and not-for-profit organisations fund work through grants, contracts and partnerships, and each may ask for different evidence of quality. Clear documentation reduces onboarding time for new staff, supports audit and grant reporting, and makes handover possible when a project ends. It also enables organisations to reuse qualitative evidence responsibly rather than treating it as disposable content for a single report.

A practical documentation standard

A workable standard can be built from two connected checklists. The first governs the research record itself; the second prepares that record for sharing. The lists are deliberately modest because a small, maintained system is more useful than an elaborate template no one completes.

  • Record identity: Give every project, site, participant, file and version a stable identifier.
  • Capture context: Note dates, location, language, method, equipment, observer and relevant conditions.
  • Track decisions: Log consent changes, exclusions, translations, code revisions and analytic corrections.
  • Separate layers: Keep source material, working derivatives, interpretations and outputs distinguishable.
  • Protect access: Record embargoes, risks, permitted audiences and the authority for disclosure.

The second checklist turns the working record into a responsible deposit. Before release, someone other than the data collector should be able to follow the path from a claim to its supporting material. Review is not a claim that every transcript is perfect; it is a check that important transformations are visible and access conditions are workable.

  • Describe the study: Supply a plain-language summary of aims, methods, participants, limitations and intended uses.
  • Explain the files: Provide a data dictionary, codebook, file relationships and guidance for non-specialist readers.
  • Minimise harm: Apply defensible de-identification, address cultural obligations and test re-identification risks.
  • State provenance: Credit collectors, translators, community advisers, repositories and earlier reuse.
  • Preserve continuity: Name a contact or custodian, review dates, formats and retention triggers.

IASSIST 2017 is most valuable when read as a record of connected practice rather than a formula to copy. Its Kansas sessions linked sophisticated research themes with the ordinary need to communicate across roles and systems. For qualitative work, the enduring lesson is to document decisions while they are made, preserve ambiguity without pretending to resolve it, and make access conditions explicit. Australian teams can apply that lesson by treating documentation as shared infrastructure for trustworthy, respectful and genuinely reusable research.

At the Conference

What attendees experienced in Lawrence

p1 reed

Plenary Sessions

Keynotes from Daniel Reed on data, technology, and culture, and Jennifer Clarke on digital agriculture and the Midwest Big Data Hub.

platinum icpsr

Workshops

Full-day technical workshops on Tuesday, May 23. Morning sessions ran 9:00–12:00 and afternoon sessions 13:00–16:00. Laptops were required.

gold rockhurst

Social Events

An opening reception at The Oread, a banquet, and an optional post-conference tour of Kansas City including Crescent Moon Winery.

Venue & Accommodations

Where the conference took place

silver ifdo

Kansas Union

University of Kansas campus, Lawrence. Main conference venue with check-in on the 4th and 5th floor lobbies.

platinum ddi

The Oread

1200 Oread Avenue, Lawrence. Hosted the opening reception and offered a room block for attendees.

silver ciser

The Eldridge

701 Massachusetts Street, Lawrence. A partner hotel with a reserved room block for conference guests.

platinum icpsr

Springhill & TownePlace Suites

Marriott properties in Lawrence with room blocks reserved under the "KU IASSIST Conference" name.

Program Highlights

Sessions and activities

Plenary Sessions Workshops Poster Presentations Committee Meetings Opening Reception Banquet Tour Kansas City Pecha Kucha Check-In Local Favorites

Getting Here

Lawrence, Kansas

Kansas Union · University of Kansas
1301 Jayhawk Blvd, Lawrence, KS 66045

Kansas City International Airport (MCI) is approximately 50 minutes by car. Lawrence Transit Routes 10 and 11 served the area ($1 exact change).

Plan Your Stay

Accommodation options that were available

silver ifdo

The Eldridge

701 Massachusetts Street, Lawrence, KS 66044. Room block now closed.

platinum ddi

The Oread

1200 Oread Avenue, Lawrence, KS 66044. Room block now closed.

silver ciser

Springhill Suites

Marriott property. Room block reserved under "KU IASSIST Conference."

gold rockhurst

TownePlace Suites

Marriott property. Room block reserved under "KU IASSIST Conference."