Exploring digital innovation in data collection
In an era where artificial intelligence (AI) is increasingly being leveraged to achieve the Sustainable Development Goals (SDGs), digital tools hold immense potential for development initiatives. However, in rural areas with limited internet access, AI-based solutions might seem unachievable.
Gathering insights from communities is vital to understand their water needs. The International Centre for Integrated Mountain Development (ICIMOD) and Frank Water collaborated to conduct household surveys in two springshed sites in Nepal’s Kavrepalanchowk district (Opi and Bhelwati) using Frank Water’s Water, Sanitation and Hygiene (WASH) Connect tool. Alongside these surveys, a pilot initiative by Colectiv, Frank Water, and ICIMOD tested the feasibility of collecting qualitative data through voice notes. This method aimed to assess the feasibility and efficiency gains from using AI to assist with transcription, translation, and analysis of qualitative data.
This case study highlights how a hybrid approach combining offline data collection with AI-supported analysis can enhance qualitative research in remote regions.
The pilot: Voice notes as a data collection tool
Community resource persons conducted 31 interviews with local householders, asking two key questions:
- Drinking water preferences: Can you tell me about the different water sources you use for drinking? Which one do you consider the best, and why? (पिउने पानीका लागि तपाईंले प्रयोग गर्ने पानीका स्रोतहरूबारे बताउन सक्नुहुन्छ? तीमध्ये कुन स्रोत तपाईंलाई सबैभन्दा उपयुक्त लाग्छ र किन?)
-
Challenges in water access: Can you tell me about any problems or difficulties you face in accessing clean drinking water? Would you like to make any changes to your drinking water source? (सफा पिउने पानी प्राप्त गर्न तपाईंले भोग्नु भएका कुनै समस्या वा कठिनाइहरूबारे भन्न सक्नुहुन्छ? के तपाईं आफ्नो पिउने पानीको स्रोतमा कुनै परिवर्तन गर्न चाहनुहुन्छ?)
Community resource persons recorded participants' responses as audio notes using mobile phones. The files were anonymised, uploaded to an encrypted online folder, and later transcribed, translated and analysed using Colectiv’s AI-based qualitative analysis tool.
Insights from the field
Feasibility and challenges
- Community resource persons successfully recorded and transferred voice notes.
- Environmental factors like harvesting seasons and severe flooding and landslide delayed data collection. This meant that we had to conduct qualitative interviews after the WASH connect survey, and this was inconvenient for both surveyors and participants.
- Background noise from people and animals and poor recording conditions affected recording quality, and this had an impact on transcription accuracy, requiring human verification for clarity.
- AI-based transcription of Nepali was functional but imperfect. The audio quality issues combined with dialect variations, and frequent use of proper nouns made a challenging test-case.
- AI-support analysis of qualitative data worked well to identify important insights from community members.
What did the community say?
Community members had a strong preference for drinking spring water. Spring water tastes sweet and good, and people feel it is healthy and good.
म ओपी मूलको पानी पिउँछु। म जन्मेदेखि नै यही पानी पिइरहेको छु। यो पानी मलाई एकदमै स्वादिलो लाग्छ। यो नै सबैभन्दा राम्रो पानी हो किनभने यो अत्यन्तै मिठो छ । यो पानी एकपटक पिएपछि अरू कुनै पानी पिउन मन लाग्दैन। (I drink water from Opi spring, I've been drinking it since birth, it feels good, it's the best water because it's incredibly sweet. Once you drink it, you do not feel like drinking any other water)
Tap water and stored tank water were alternatives for a few people, but many avoided these sources. They use this water only for livestock or washing.
हाम्रो बोरिङ पनि छ, तर बोरिङको पानी पिउनको लागि त्यति योग्य छैन।लुगा धुन, भाडा मोल्न मात्र प्रयोग गर्ने गरिएको छ। (The water from our boring well is not as suitable, and it is only good for washing clothes and utensils.)
The main challenges they face are that the spring source can be far away and, especially in the rainy season, paths can become slippery and impassable. The spring can also get contaminated with overflowing water.
बर्खामा बाटो चिप्लो हुन्छ र हिउँदमाआफूलाईचाहिएकोजति पानी पाइदैन। (During the rainy season, the paths get slippery, and sometimes it's not easy to get as much water as wanted in winter season.)
बर्खामा मूलको पानी धमिलो हुन्छ, कहिलेकाहीँ किराफट्याङ्ग्रा पनि जम्मा हुन्छन्, र सफा पानी पाउने कुनै सम्भावना हुँदैन। (During the rainy season, it becomes muddy, sometimes insects accumulate, and there’s no way to get clean water)
People wanted improvements in infrastructures to help them have better drinking water access. Many people requested that their preferred spring water be brought closer to their homes to reduce the burden of collecting it. If the water could be piped directly to households, or at least to nearby tanks or reservoirs, it would help avoid the difficulties of collecting water along muddy paths.
ट्यांकी बनाइदिएर धारो जडान गरिदिए सजिलो हुन्थ्यो।बूढाबुढी बारम्बार पानी ल्याएर खान सक्दैनन्। (If a tank is built or a tap is provided it would be much easier. Elderly people can’t keep carrying water back and forth)
हरेक घरमा धारो जडान गरिदिए कस्तो सजिलो हुने थियो, मलाई त यस्तै लाग्छ! (If taps could be provided at every house, it would be convenient, that's what I feel)
Others requested improvements to existing sources, such as better pathways and protective measures to prevent contamination and overflow.
पानी नपस्ने गरी अलिकति ढलानसहित पर्खाल बनाउने र सम्भव भएमा वरिपरीको भुइँ पनि ढलान गरेर ढोकाहाल्नसके अझ राम्रो र सुरक्षित हुने थियो। (It would be better and safer to construct a wall with a slight slope so that water doesn't enter, and if possible, to install a door with the surrounding ground sloped accordingly.)
पँधेरोमा जाने बाटोअलि राम्रो बनाइदिनु पर्छ।पँधेरो वरिपरी खनेर आसपासको क्षेत्र अलिकति ठूलो बनाइदिनु पर्छ र राम्रोसँग संरक्षण गरिनुपर्छ। (It would help if the roads were improved. The area around the source needs to be a little bigger and better maintained)
Community partner reflections on the use of an AI tool
As a partnership, we reflected on the inclusion of an AI tool for data collection and remote analysis. The process of solving development problems in remote regions across world has always meant involving ‘people from outside’ these communities. How much ever one may try – it’s difficult to bridge the gap between what is communicated by communities and what is understood by the ‘people from outside.’ The use of a tool like Colectiv reduces the communication gap drastically as the interpretation of what is spoken does not lie with individuals recording their responses, or in the reduction to survey items. Instead, this tool allowed all the subjective answers community members provided to be recorded verbatim without any interpretation and shows (qualitatively and quantitatively) what the community members want across various demographics within the community. We feel that needs assessment and understanding of community perceptions from such tools is closer to what the community means and not what is in the heads of interpreters, or those selected involved in the project.
Key takeaways
This pilot demonstrated that AI-driven transcription and translation can support qualitative data collection in remote communities. In this case, human oversight was important for accuracy, but this may be less essential in other contexts. More broadly, the use of voice notes enabled researchers to capture community-driven narratives, providing valuable insights beyond quantitative survey responses.
Recommendations
- Expand the use of audio notes: Integrating qualitative questions into existing surveys can be a cost-effective way to amplify community voices.
- Explore minimum digital solutions: There is value in leveraging technologies that are available in the field in hybrid approaches.
- Improve AI capabilities in Nepali: While AI transcription is viable, more development is needed to handle dialects, accents, and non-standard pronunciations effectively.
- Invest in water infrastructure: Community members have expressed a clear need for household taps, storage solutions, and improved water access routes – critical investments to enhance water security in the region.
By combining digital tools with human-centered approaches, organisations like ICIMOD, Frank Water and Colectiv can enhance development outcomes by ensuring that community perspectives can be used to improve programme planning and delivery. The value of a household survey can be greatly increased by adding tools that gather peoples’ voices and put them at the heart of decision-making.
