Are you interested in what we are currently working on? Then you have come to the right place!
Welcome to our "Working Out Loud" page! In addition to our regular blog posts, we want to offer you a glimpse into our current activities. The idea is to foster collaboration, build connections and accelerate learning. Here, we'll share summaries and insights of our concepts, ideas, thoughts and ongoing activities that are either in the developmental stage or still in the exploratory phase.
While our blog posts dive deeper into our established processes, this page is a place where we can present our work-in-progress and invite your feedback. Some of the topics we present here are already in advanced stages and have corresponding blog posts linked to them, while others are still in their early stages of development.
As Data Lab, we strongly believe in the power of collaboration and are excited to engage with you on our ongoing journey. If you're interested in any of our activities or have some thoughts, ideas or even criticism, we'd love to hear your insights and feedback. Make sure to contact us and join the conversation. Let's work out loud together!
Contact us at: datalab@giz.de
AI tool for UN Plastic Treaty
Every year, 430 million tons of new plastics are produced, and this amount could triple by 2060. Sixty percent of plastics last less than five years and only nine percent are recycled. Microplastics have entered our bodies, our water, our air and our soil. In March 2022, the UN Environment Assembly adopted a historic resolution for a legally binding agreement on plastic pollution and aims to conclude negotiations by the end of 2024. In each round of negotiations, delegates from over 160 countries express their positions in the form of texts or oral statements, which are then documented. The current process requires the INC (Intergovernmental Negotiating Committee on Plastic Pollution) Secretariat to manually search through large amounts of text, which is time-consuming and prone to errors.
To support the INC in the negotiations, the Data Lab and the GIZ Project Marine Litter Prevention are developing an AI tool based on RAG (Retrieval Augmented Generation), which makes it possible to chat with the data to receive automatically generated answers and links to the relevant documents.
Preparations for INC-5 /
Following the last negotiation round in April 2024, we are now gearing up for the expected final round of negotiations in November 2024 (INC-5). Together with our partners from the GO Circular project and the BMUV Plastic Treaty Support Programme, we have established guidelines and parameters for the next update iteration ahead of INC-5. In the lead-up to the negotiations, we will focus on intensive feature development to enhance the app's user-friendliness and expand its analytical tools.
First Prototype /
We have successfully created a knowledge base derived on the relevant documents and submissions in the UN Plastic Treaty negotiations. This knowledge database serves as the basis for the rag pipeline, which we have effectively implemented. This breakthrough has enabled us to create the first publicly accessible prototype of our chatbot, currently undergoing testing for quality and usability.
AI Image Recognition for Electronics Waste in Ghana
Driven by rapid technological advancements and the growing consumption of electronic devices, electronic waste is becoming a problem of global dimensions. The WHO documents that only 22.3% of the annual e-waste is formally collected and recycled. To tackle the e-waste crisis, we are collaborating with local stakeholders in Ghana to develop an AI-based image recognition application. The app is designed to assist users in the identification, categorization and documentation of electronic waste. By empowering local authorities, waste management organizations, and recyclers, our project aims to identify health and environmental risks, promote sustainable practices and contribute to a circular economy.
E-Waste Kick-off /
We are excited to share our newest Experiment! Partnering with local stakeholders in the Ghanaian recycling industry, we devised a plan to develop an advanced AI image recognition, which shall be integrated into local material flow processes. We imagine an application for handheld mobile devices, which allows for automatic scanning of landfill intake. Besides harmonizing stocktaking processes, working conditions of local waste collectors shall be improved by giving information about toxicity and value of different EWaste classes. Stay tuned!
DPPD - Road To Implementation
After the successful implementation of the DPPD Masterclass in November 2022, some teams are on their way to implement their proposed use case ideas. Topics of some of the projects deal with proven practices for reduced deforestation in Ghana or resilient wetland conservation in India. The Data Lab offers assistance with personal support with regards to technical aspects as well as deeper methodological DPPD insights and project fine-tuning. Furthermore, a self-paced atingi course on DPPD applications is in development.
Final Steps for Atingi Course /
Nearly two years after the instalment of the DPPD Masterclass, we are pleased to announce that a self-paced online course, "Getting Started with Data-Powered Positive Deviance," is nearing completion. This course will allow learners to explore the stages of the DPPD methodology at their own pace.
Field Trips Completed /
For the DPPD Use Case in Ghana on sustainable forest management, together with colleagues from the REACH project as well as local consultants, we have successfully completed approximately 20 field trips, visiting both positive and negative deviants. Through peer group discussions, site visits, and expert interviews, we are now eager to analyze the transcriptions and extract insights into sustainable, previously undiscovered practices.
Ghana Use Case /
We're thrilled to share our latest blog post highlighting our Ghana Use Case. In this post, we provide an in-depth overview of the background of this project and on how we integrate satellite data with various stages of the DPPD method. Please check it out:
Ghana Use Case /
Advancing through the stages of our Ghana use case, we've now entered the third phase of the DPPD method. Initially, we pinpointed positive deviants among communities in North Ghana, maintaining or increasing tree cover. In the subsequent stage, we utilized remote sensing data to categorize and identify these communities based on fluctuations in the Normalized Differential Vegetation Index (NDVI) over the span of 2018 to 2023. As we prepare for stage three, the fieldwork, we've conducted a preliminary validation process using satellite imagery to assess potential factors contributing to positive deviance amidst deforestation pressures in the project area. Stay tuned for further developments.
DPPD Webinar /
On May 10, the GIZ Data Lab, the Sector Network Natural Resources and Rural Development in Asia and the Pacific (SNRD AP), and Sector Network Rural Development Africa (SNRD Africa) hosted a one-hour webinar on Exploring Innovative Digital Solutions for Tackling the Climate Crisis: Insights from the DPPD Masterclass. The use of digital data in development cooperation through the innovative DPPD methodology was presented and discussed in detail. In addition, two teams from Ghana and India who are actively engaged in implementing their use cases presented their insights on data-based projects on deforestation and wetlands reservation.
Update /
In case you missed the activities around the DPPD Masterclass we hosted in November 2022, here you can find a summary in form of a live blog with key findings, interviews and explanatory texts on the DPPD methodology and some use cases that emerged from the Masterclass.
Data Feminism
After the successful Data Feminism Event series conducted in partnership with the Data Pop Alliance, the Data Lab is working on concrete possibilities to implement a feminist approach to data in GIZ's partner countries. In close cooperation with think tanks like Data2X or the aapti Institute and inspired by Catherine d'Ignazio and Lauren Klein, authors of the book Data Feminism, the Data Lab is benefiting from the momentum of the German feminist development cooperation recently launched and explore how it can bring gender data approaches to the next level.
AI-assisted Feminist City Planning at the APTF /
GIZ Data Lab has successfully completed a small-scale experiment together with the TUMI Initiative and the NDC Transport Initiative Asia, using generative AI to facilitate feminist city planning and spark discussions around more inclusive urban sites. Read more about it in our Blog Post.
Event Series and Gender Data /
GIZ Data Lab’s Event Series on Data Feminism is featured in the new Data2X Global Synthesis Report “Building Political Will for Gender Data”. The report compiles research from 30 countries, mapping the environment around the collection, analysis and use of gender data at national level in order to develop effective advocacy strategies for greater prioritization and resourcing for gender data.
Find out about our event series as an effective tool to build capacities to guarantee political support on page 4 of the report.
See the presentation of the report in this event hosted on CSW68 by Data2X and the Office of the Women's Rights Advisor to the President of Kenya.
Practices /
As of March 24, we are exploring different implementations of Data Feminism in practice: a data-feminist curriculum for university students in India, Data Feminism measures in city planning in Bangladesh and many more. Want to share your ideas for a data-feminist experiment? Get in touch!
Unveiling Vulnerabilities in Climate Policy
The global climate crisis affects everyone, but particularly impacts disadvantaged and vulnerable groups, such as children, the elderly, coastal communities, and the LGBTQI+ community, due to their socio-economic status. It's essential to integrate their concerns into national climate policies for effective and equitable climate action. However, analyzing extensive policy documents poses a significant challenge. To address this, the CPV pilot uses natural language processing (NLP) to identify and list contextual mentions of vulnerable groups, enabling gap identification and recommendations. It also incorporates a chat-based AI assistant, fine-tuned for user questions related to vulnerable groups in climate policy documents. Currently, the database covers 10 sub-Saharan African countries, starting with Kenya.
We have been selected for the UN World Data Forum! /
With our "Climate Policy Vulnerability Tracker" app, we were selected from over 600 submissions for a demo session at the UN World Data Forum in Medellín, Colombia. Before the WDF starts in mid November 2024, we will continue to develop our app, add new features and finalize a case study report for all African countries (currently the report covers 10 selected countries).
App release /
After successful prototyping with our partners in Kenya, we have released the final app version of the “climate policy vulnerability tracker”. It can be used by all interested users on Hugging Face under the following link. But we don't want to stop there: We are planning an interactive report that will initially take a closer look at selected SSA countries and then, towards the end of the year, examine all current African NDCs for demographic groups in vulnerable situations and prepare them visually - stay tuned!
Vulnerable groups /
In order to encompass a wide spectrum of vulnerable groups, we departed from our initial project concept, which centered around "youth and gender." Instead, we expanded our scope to encompass over ten distinct categories of vulnerable populations. We searched through numerous climate policy documents, identifying relevant text segments, and profoundly associating them with the relevant groups. Following extensive refinement of the classifier, it now consistently and accurately categorizes text segments into their respective groups in nearly all cases. This robust domain-specific classifier serves as the cornerstone for future generative AI application possibilities.