Beyond Protocol: Reflecting on the Sela Platform Mid April 2018 Pilot
The Sela Platform was initially conceived to be used in the context of project verification in the Niger Delta. It was created to improve transparency and accountability between public service beneficiaries and those acting on their behalf. It is composed of three main modules.
The first module is a web application where project funders can create, read, update, delete projects. The second module is a conversational interface engaging with evaluation agents to understand the real time status of projects. It takes input from contractors and evaluation agents. The last module is a wallet module rewarding community members for their observation using cryptocurrency.
During mid April, the Sela team got to perform a real-world user test of the platform in the context of an oil cleanup performed by Sustainability International, an NGO focused on raising environmental awareness and catalyzing cleanups in the Niger Delta. The main learnings include that community members were comfortable interacting with a conversational interface and a cryptocurrency wallet; however, there is room to simplify the mobile wallet experience if the goal of cryptocurrency is to reach the unbanked, who are also often illiterate.
Introduction
The Niger Delta has a population of more than 31 million people whose life expectancy is just over 40 years old. The region’s oil revenues account for 85% of the Nigerian government’s budget, yet is plagued by more than 7,000 unresolved oil spills. Several factors contribute to this unstable dynamic.The first one is lack of accountability. Even well-meaning projects to provide basic human needs for the people of the Niger Delta are routinely abandoned. Over $1B in remediation funds have flowed to the region since 2012, yet only 12% of projects were executed. Vendors tasked with cleaning up oil spills often collude with government officials, so funds allocated to cleanup projects are not used appropriately. Oil spills have ruined the once-fertile local terrain, seeping deep into the soil and damaging the region’s agricultural economies. This has upended the economic stability and even the health of many of the region’s workers. As a result of systemic misdirection of funds, the substantial oil profits do not “trickle down” to local communities. Compounded by damage to the farmland, Nigerians have a much harder time making a livable income.
Recent technological progress offer the avenue to provide technical solutions to the challenge of achieving public service transparency.For the benefit of these technologies to be realized, there are several aspects of the cryptocurrency user experience which will have to be simplified. Particularly if the goal is to enable unbanked and illiterate users to send transactions logged on the blockchain.
1) Modules
The MVP of the Sela platform has three modules. The first one is a web application allowing project funders to create, read, update and delete projects. The second one is a conversational interface allowing evaluation agents and contractors to mark tasks as complete and provide evidence of completion. The last one is a mobile wallet where community observers can receive tokens for sending their observation. The Sela protocol is an abstraction layer handling assigning a reputation to each actor, triggering payments, creating and resolving disputes.
- The web application is a dashboard using the model, view, controllers architecture with the frontend components written in React and the backend being a NodeJS server. In this MVP, the web application enables project funders to sign up, login, create projects, tasks and review a timeline of evidence provided by contractors to track projects throughout their execution.
- The texting platform is a conversational interface taking input from community members around the execution of tasks. These input include geotagged picture or videos as well as answers to tailored, project-specific questions. For example, if the project funder wants to build a school which will contain three buildings with each of them having 2 stories and 4 classrooms per floor, the observation sent by evaluation agents will include geotagged pictures/videos as well as answers to questions such as “Are there 2 stories per building ?” ; “Are there 4 classrooms per floor?” or “Has the fence been built ?”. In the MVP implementation, the conversational interface takes the form of a Telegram bot named “Selabot”. Selabot will only talk to registered Sela users and once a conversation is initiated, it will ask questions based on the task the conversation has been initiated for as well as the current status of projects. This module directly connects citizens and make their input visible. Before, funders had no way to connect with beneficiaries at scale and a chatbot enables such interactions, providing real-time feedback on state of projects during and after implementation.
- The last module involves the financial interactions powered by cryptocurrencies. In the current implementation of the MVP, it takes the form of an open source Stellar wallet called “Lumina” to which cryptocurrencies are sent upon receiving the data from evaluation agents. Lumina runs on the Stellar network and enables users to create their own wallet which becomes password protected so that instead of asking for private key during a transfer, only a password is asked. For this test, we also created a Sela token on the Stellar Lumens. So an extra step involved configuring the wallets to trust the account disbursing the Sela tokens.
We chose to do a progressive decentralization approach as can be seen in our tech stack by gradually integrating IPFS and decentralized ledgers as we iterate between having a functional product and sharing data about ongoing projects.
2) User testing
A.Scope
For the Mid April pilot, the Sela team wanted to stress test a few assumptions and conducted user testing. The user testing was broken down in two parts: The first one involved the ability to submit geotagged and timestamped data about project outcomes and the second one involved ability to interact with a cryptocurrency wallet. The assumptions tested during this user study involved ability to upload data, answer questions around the project development, receive cryptocurrency and transfer the cryptocurrency to a liquidity point who would exchange it for the local currency.
For the pilot, we provided the smartphones and already preconfigured them by installing telegram, creating generic contact names and started a conversation with the Selabot contact and also setup the wallets and password for each user. We made this choice to narrow down the scope of the assumptions tested and just learn about user’s abilities to interact with an already set up wallet and conversational interface, assuming that onboarding can be worked out. The location of the user test was in Gokana, a rural area of the Niger Delta and the first part involved 6 users including 4 women and 2 men and the second part involved 4 users including two men with at least a high school education and two women, one having at least a middle school education and the second one having no schooling.
B. Tasks to be performed
The tasks to be performed included answering multiple-choice and domain-specific questions, uploading pictures/videos through a conversational interface, receive tokens and transfer them to the liquidity point. Transferring tokens included opening the Lumina wallet, scanning the QR code of the recipient, typing the amount of tokens to be exchanged, including a memo and entering a password. In this case, to simplify the assumptions tested, the password was known in advance to just focus on the user experience around the transfer of tokens. In addition, users were shown each of the steps by a trainer. The questions involved selecting the project for which they were reporting, selecting the task for which they were reporting, taking a video or a picture as well as answering whether the task got completed.
3)Results
A.Data upload
For the first part of the user testing which involved answering questions and uploading data, all 6 users were able to answer questions and take video/pictures and submit them through the conversational interface. The average time required to complete the task was 45 seconds.
The most confusing part and area to be improved involved picking the task for which users were reporting. The current project was broken down into 11 tasks and the multiple choice format was long to read and the most confusing part. Other ways to improve the current experience involve either automatically picking a task for which user report or include a small subset of the tasks based on previously collected answers.
Another area for improvement include using emojis as cues for instructions. For example, instead of asking whether a task was done through Yes/No question. A thumbs up/thumbs down menu can be created. Similarly, when asking to upload picture, an emoji representing a camera can be included in the prompt to create visual cues for the user.
B.Token transfer
This second part of the user testing involved four users.In this user test, the metrics we measured included time for the user to perform a token transfer, the time for the transfer of token to become visible to the liquidity point and the number of users who were able to successfully transfer tokens.Three out of the four users were able to transfer tokens under five minutes and the user who was illiterate could not spell the provided password.
The median transfer length including scanning the QR code, typing the token amount, including a memo and providing the password to validate the transaction took a median time of 80 seconds. Some of the challenges observed included extra time to scan the QR code, network issues that made some transactions having to be submitted again.
The median time it took between the token transfer and the availability of the transfer on the wallet of the recipient was 12 seconds. This takes into account the time it takes to validate the transaction on the Stellar network and refresh the recipient’s wallet by polling the network and receiving the update.
The speed of the Stellar network given its different consensus protocol (federated byzantine agreement) leading to fewer nodes was taken into account when performing this real world test. The needs were close to the speeds required for retail use cases of cryptocurrencies.
For fun here is a sample token transfer done by Christy during this testing period : https://steexp.com/tx/865c7b1440771d63aeecd428ea4b959ce5bbf095b19dd05d925a7ee2a900c548
4) Discussion
This real world user test reveals that the current experience available for public key and private key management for cryptocurrency is not user friendly enough to enable illiterate users to transfer cryptocurrencies. While this may seem negligible, given that in countries such as Switzerland, the literacy rate is above 98%, in other countries such as Ivory Coast or Nigeria, the literacy rate hovers round 50–60%.
If the promise of cryptocurrencies is to be used as a worldwide payment system and help “bank the unbanked”, then more steps will have to be taken to simplify the user experience. In addition, thinking about simplifying the public/private key management experience may also benefit users.
The current user experience is analogous to carrying all of one’s money in a safe, which needs to be carried and unlocked each time one wants to make a payment. While safes guarantee the maximum security, over time humanity has developed several mechanisms offering different security and user experience tradeoffs including savings account, checking account, credit cards, debit cards and physical wallets used to hold cash.
To fit the needs of users at different comfort levels with manage their public/private keys, such analogies will need to be developed to further mainstream adoption of cryptocurrencies. In the following section, we will discuss three options and analyze the various security/user experience tradeoffs.
5) Potential solutions
A. Low security wallet
One of the most radical solutions maximizing user experience and minimizing security is to have a low security wallet that is not password protected. It is analogous to carrying cash in a personal wallet. Restrictions to reduce security tradeoffs can include limits on the amount of tokens the wallet can hold and the number of transfers per day. The analogy is that in real life few people carry more $2000 in their everyday wallets.
Such a wallet can be coupled with a more high security wallet which can transfer it a fixed amount per month or require a password before transferring to the low security wallet.
B. Emoji and Images as passwords
A second option to simplify the user experience when initiating a transaction can include having the user pick an image as its secret image. In this example, a wallet would have a gallery of images and the user would pick one of them as its image. Each time a transfer is to be initiated, he can, for example, be asked to pick his image or answer whether or not his image is present among options presented. A more interesting idea can be for a user to remember a key concept and randomly generate or retrieve an image including that concept and others which don’t. A practical example can be that the user “secret code” is horse and when sending a transaction, one image containing horse can be retrieved (or generated) and others not containing this concept can be generated and the user would just have to pick an image including this concept.
C. Speaker and speech recognition as passwords
The last option can include having the user say a phrase and pick it as a password and provide additional vocal samples. Each time a transfer of token is to be made, a machine learning algorithm can assert whether the speaker registered with this wallet uttered his passphrase. The advantage of this approach is that it is friendly for illiterate users assuming high speaker recognition rate and low word error rate while guaranteeing almost the same security as a password protected wallets. Further research and experimentation on combined speaker + word recognition will inform the adoption of this technology.
Conclusion
To conclude, over the last decade, progress in cloud computing, artificial intelligence and distributed ledger technology offer us the promise to hold government accountable and usher an era of unprecedented prosperity across the African continent. Imagine a world where conversational and voice interface enabled governments to scalably engage with their citizens to monitor progress on projects, public sentiment and where public spending and outcomes were shared on a distributed ledger with citizens. A lot of research, engineering, business, political and design efforts have to be made to make that future a reality but at Sela that’s where we think the world is headed and we are passionate about creating the digital infrastructure powering this future.
Beyond Protocol is a blog series by Sela taking readers through the journey and thought processes involved in using cutting-edge technologies to drive tangible impact in the world
Hassan Kané is a recent graduate of MIT. He’s a tech entrepreneur and AI researcher currently serving as the CTO of Sela.
Sela is a technology company building the digital infrastructure to create communities of transparency in the emerging world.
