Artificial Intelligence

Using Ethical AI To Turn Citizens’ Information Into Insight

Citibeats AI data

In the service of business, of society at large, artificial intelligence (AI) can be effective. It can be fast.

Can it also be ethical?

The wisdom of crowds, gleaned from social media, can paint a gestalt picture of how a government agency’s, bank’s or retailer’s efforts are being received on the ground, so to speak. And it can also (perhaps), fed through models and analytics, can bolster decision-making for the greater, common good.

Public opinion matters, after all, but across the social media platforms, the chatrooms —  the chatbots, even — making sense of qualitative data is a challenge for most enterprises.

The review, the opinions, thumbs up and down, and information passed in 240-character bytes are too far-flung, and constant, to monitor effectively.

To that end, according to Citibeats CEO Ivan Caballero, text analysis — done at scale — can help pinpoint and solve challenges before they mushroom into problems.

The Citibeats platform (based in Barcelona, Spain), he told PYMNTS, analyzes and offers insight tied to natural language data, sourced from social media or transcripts from chatbots, for example.

Said Caballero, the connected economy — spanning a plethora of channels — benefit from such insights.

Ethical AI

Underpinning it all — the data collection (across public forums or paid sources, such as through Twitter premium application programming interfaces (APIs), or proprietary data from Citibeats clients), analysis and the insight derived from that – is the concept of “ethical AI.”

Accessing and ethically processing data, Caballero said, means ensuring the privacy of all individuals who provide that data. There must also be transparency, as the enterprises and agencies using the platform must have obtained end-users’  explicit consent to work with that data. Bias, as might be seen across geographies or demographic groups, must be eliminated and the data must be anonymized, he said.

As Caballero said, “we can use the different types of data and can vary our models by mixing all of them.”

The Use Cases

To get a sense of data in the service of the greater public good, Caballero told PYMNTS that Japan, which has a long history of grappling with natural disasters, has enlisted his firm and NTT Data to analyze citizens’ responses across social media (such as TripAdvisor), using machine learning algorithms, to see where critical infrastructure repairs might be needed across several cities.

Other use cases: Spain’s Navarra government created a model that can scour Instagram, Facebook and other sources to pinpoint hate speech and help shape educational and local government efforts to protect vulnerable communities.

In a larger setting, where global events such as the coronavirus have both a health and social component, Caballero maintained that platforms including Citibeats could help provide insights so that health organizations (among other stakeholders) can react with greater speed and efficiency. The collected intelligence, he said, can provide clarity into what he termed “changes in behavior” in areas where the virus is taking root.

The Commercial Setting

With a nod toward commercial use cases, Caballero noted, too, that banks looking at financing risks — say for restaurants — can gain insight from trends such as tourism, or review of delivery platforms.

Social media-sourced data can help financial services firms, and regulators, improve financial inclusion efforts and customer service, which in turn boosts customer loyalty.

“Qualitative data is difficult for banks to manage,” said Caballero, “and what we’re doing is helping the banks define a gap in their services, especially those provided to poorer people.”

Caballero said that Financial Sector Deepening (FSD) Kenya has been monitoring the tweets and chats tied to Kenyan financial firms, looking at the behaviors and service levels of banks’ agents across a variety of transactions. FSD Kenya is monitoring whether consumers can access their accounts or are having technical difficulties, or whether scams are impacting customers, which provides an early warning of sorts for the financial institutions (FIs).

“It’s clear we are becoming more connected, and that we are generating a lot more data. And it’s been demonstrated that people are happy to share that data if there is a direct benefit to them or for humanity in general,” he told PYMNTS.



New forms of alternative credit and point-of-sale (POS) lending options like ‘buy now, pay later’ (BNPL) leverage the growing influence of payments choice on customer loyalty. Nearly 60 percent of consumers say such digital options now influence where and how they shop—especially touchless payments and robust, well-crafted ecommerce checkouts—so, merchants have a clear mandate: understand what has changed and adjust accordingly. Join PYMNTS CEO Karen Webster together with PayPal’s Greg Lisiewski, BigCommerce’s Mark Rosales, and Adore Me’s Camille Kress as they spotlight key findings from the new PYMNTS-PayPal study, “How We Shop” and map out faster, better pathways to a stronger recovery.