Deeper Understanding: Addressing Methodological Constraints and Ethical Implications of Humanitarian Needs Assessments Using Natural Language Processing
Bibliographic record
Abstract
Humanitarian needs assessments (HNAs) are essential for prioritizing and informing the provision of humanitarian assistance. Data for HNAs are commonly collected using computer-assisted personal interviewing (CAPI) tools. However, the transcription and translation of qualitative interview responses during a humanitarian crisis is challenging to conduct rapidly and at scale. This has led to calls for deploying natural language processing (NLP), a type of artificial intelligence (AI), to enhance CAPI tools. Despite this, data on current methodological HNA practices and issues, as well as the risks to affected populations or organizations providing humanitarian assistance due to data processing, including AI use, are not well understood.\n\nThis manuscript-based study employed a three-stage mixed-methods Design Science Research approach to inform the design and empirical evaluation of new features for KoboToolbox, a CAPI tool supported and used by humanitarian organizations, to systematically transcribe and translate answers to open-ended questions (OEQ) from HNA interviews. The first stage involved conducting interviews with international key informants (KIs) to explore current approaches to HNA and constraints related to different interview methods. Findings from interviews with 23 KIs from 13 countries revealed 47 themes.\n\nThe second stage comprised a scoping review to map the range of ethical issues raised in academic literature concerning the processing of data from people affected by humanitarian crises. The scoping review identified 100 relevant studies after screening 8,387 papers, resulting in 22 themes. In the third stage, new KoboToolbox software features were designed and implemented based on a user-centered design approach, with usability testing conducted using observations and feedback from test users.\n\nThemes from stages 1 and 2 were used to inform the iterative design and software implementation process, which was validated through feedback from 14 test users from 10 countries. User testing demonstrated strong support for the applicability and usability of the new features, and the results were used to identify and address several usability issues in the software. The tool resulting from this research meets professional requirements and can now be readily deployed by humanitarian organizations worldwide to systematically transcribe, translate, and analyze answers to OEQs from HNA interviews.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.448 | 0.605 |
| Meta-epidemiology (narrow) | 0.001 | 0.002 |
| Meta-epidemiology (broad) | 0.003 | 0.002 |
| Bibliometrics | 0.016 | 0.013 |
| Science and technology studies | 0.007 | 0.020 |
| Scholarly communication | 0.030 | 0.033 |
| Open science | 0.006 | 0.021 |
| Research integrity | 0.008 | 0.008 |
| Insufficient payload (model declined to judge) | 0.004 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".