Data is a double-edged sword. On one hand, we need the numbers in order for government and institutional leaders to develop and, ultimately, implement programs for law and policy change. On the other, such a narrow focus on the numbers trivializes the human beings behind it all – which is who the data is supposed to be representing in the first place.
Human data in our technology-driven world lives in the form of text, not numbers; a barrier that’s hindered the development of a more responsive and abundant society.
We see AI and machine learning step in and improve so many aspects of our daily lives and in the areas of banking and personal finance with mobile check deposits and fraud prevention alerts; in our mobile phone use with automated transcription for voice-to-text capabilities; in our online shopping experience with product recommendations on Amazon. And the list goes on.
With everything AI can do, what if it could help us break this human/data barrier? Just imagine: citizens could be the sensors for a more responsive society.
Can Machine Learning Make Data More Human?
The short answer to this question is, yes.
Citibeats machine learning technology gives a human voice to nonhuman data. It works like this:
First, it identifies insights and trends by collecting information – in this case, text – across all networks in any language. The text gathered is all-inclusive, from all social media channels – Facebook, Twitter, Instagram, LinkedIn and such – to chat forums, blogs and any other relevant platform where citizens are expressing their thoughts and opinions.
Next, proprietary machine learning algorithms go through the process of understanding the information. The AI technology sorts and structures the information into specific categories that need to be measured.
This is key because it allows decision makers to analyze relevant information for understanding the real-world conversation. This goes way beyond the numbers and the figures – it’s as if the data can now speak in a unified voice; in a voice that leaders can hearand understand.
Challenges behind human data
When you enter the realm of human data, you inherently tap into the area of personal information and privacy – an area that’s very ethically sensitive. There are complex challenges that arise when dealing with the public and people-based research. Apart from the ethics part of it all, there are also the following to consider and work to overcome:
- The availability of quality data (or lack thereof)
- Generating inclusion – data that’s not discriminatory from the onset
- Avoiding individual profiling
- Sorting out fake news
- Counterbalancing online hate speech
At Citibeats, we have practices in place to help offset some of these challenges. Some of these practices include:
- Anonymizing data
- Aggregating data to provide insights on cohorts of citizens, minimizing the effect of profiling and discriminatory skews
- Using an advanced categorization system to sift through fake news or bots
- Supporting projects that use our system to identify and counteract hate speech with education budgets
Case Studies of AI Text Analytics At Work
All of this is well and good, but doesn’t help us much if the technology isn’t producing results in the real world. So, what have been the results so far?
One of the biggest wins has been the increased speed of which data has translated into action:
- 5,000 damaged infrastructure reports from the 2018 Japan floods were extracted 21 days earlier – structuring reports of damage that led to more efficient prioritization of repairs needed across the country.
- 7,000 consumer complaints in Kenya were analyzed 45 days earlier – identifying reports of bank scams that led to timely consumer protection investigations.
- 3,000,000 citizen voices were reported 60 days earlier – quantifying public feedback on social issues that led to budget allocation according to the citizens’ priorities.
But let’s dive in and take a closer look at how this has played out and made an impact so far in the real world…
Case Study #1: Social Policy Change in the United Kingdom
The Milton Keynes Council – the local council of the Borough of Milton Keynes in Buckinghamshire, England – was shifting their approach of policy making to be more data driven, and developed MK Insight as a singular hub to collect all open data for decision-making efforts. They knew they wanted to include social media in the data set, but they were discouraged by the ever-changing and messy world of social media. How do you pull actual data from chaos?
This is where Citibeats came into play. It was set up to measure real-world dialogue about MK’s main areas of focus for pressing social issues: homelessness, drug abuse and exclusion. Citizens living around the victims of these issues were being used as sensorsvoicing their problems.
In February 2018, there was a spike in the community’s dialogue regarding concern for how the large number of homeless people in the city were going to be affected by sub-zero temperatures – an alert which the AI system detected and reported. The next three days of freezing weather conditions resulted in a national scandal that got reported by newspapers nationwide.
In this case, the citizens acted as a “social early warning system” and allowed MK Insight to discover several grass-roots organizations in the area to help tackle the issue in a timely manner. This is a real-life example of how AI can bring social issues to the forefront that may otherwise go unnoticed because they’re not easy to measure in hard numbers.
Case Study #2: Financial Inclusion and Consumer Protection in Kenya
Kenya is the trailblazer in today’s world of mobile money and digital credit. As financial services became more digital – so did consumers’ method of complaints and reports of abuse. It was impossible to manually process, let alone address, thousands of complaints that flooded in daily in an effective, time-sensitive manner. How could banks and regulators use this valuable data to flag investigation-worthy issues in a real-time way that would actually allow them to protect consumers?
Here is where Citibeats entered the scene. Financial Sector Deepening (FSD) Kenya teamed up with the Dignity and Debt Network to define the framework for which topics qualified as investigation-worthy – events like scams, blacklisting and unfair charges. From there, Citibeats AI was trained to recognize these topics and alert regulators when red-flag patterns were detected.
Not only did FSD Kenya report that the Citibeats alerts received were 80-90% relevant for investigation-meriting issues, but they found that consumer complaints were flagged 45 days earlier than before the technology was put in place. So valuable was the information, that FSD Kenya is working with a one-year complaints analysis to make humanized datarecommendations to the entire country’s financial services industry.
Download more case studies here:
- STEMcat: Promoting Science and Technology Among Youngsters
- AI Analytics for Tourism
- Measuring SDG #9 in Sant Cugat: Innovation and Infrastructure
So far, governments and countries in over 30 countries worldwide are applying Citibeats technology to rethink and restructure their strategy about using civic feedback data for actionable data to guide change. Over 10 million voices have been represented for what is now being considered in the decision-making process – which is quite a remarkable improvement from the drastically under-utilized citizen feedback in the past.
We’ve seen the impact in the areas of sustainable development, financial services, natural disaster response, social policy and hate speech policy. And we’re just scratching the surface.
Over the next three years, our goal is to influence $2 billion in budgets for efforts that have not previously taken into account the voice of the citizens. By giving a face – and a voice – to the numbers, we bring humanity back into the equation. A factor that has been ignored for far too long.
Our ultimate objective is to keep building a responsive society on a global scale – one voice at a time.