"How Fair Is AI? A Look into KakaoT's Dispatch System"

[Asia Economy Reporter Buaeri] As the taxi industry raises suspicions of 'Kakao call (dispatch) favoritism' day after day, curiosity about Kakao Mobility's artificial intelligence (AI) dispatch system is also increasing.

The four taxi organizations (National Taxi Labor Union Federation, National Democratic Taxi Labor Union Federation, National Private Taxi Transport Business Association Federation, National Taxi Transport Business Association) submitted a petition to the Fair Trade Commission on the 10th, alleging that "Kakao Mobility is abusing its monopoly position and hindering fair trade." (Refer to our exclusive report [Exclusive] Four Taxi Organizations Target Kakao... Petition to FTC on "Call Favoritism" Suspicion)

For example, when a passenger calls a taxi using the taxi-hailing application 'KakaoT,' calls are not assigned to nearby taxis, but are repeatedly assigned to 'KakaoT Blue' taxis located farther away. 'KakaoT Blue' is a brand taxi operated by Kakao.

Regarding these suspicions from taxi drivers, Kakao Mobility drew a clear line, stating, "There is no artificial dispatch as the AI analyzes the data."

A Kakao Mobility official explained, "Calls are not based simply on distance but on the taxi's 'estimated arrival time' and big data analysis is simultaneously applied to factors such as ▲driver ratings ▲driver dispatch acceptance rate ▲driver driving patterns ▲real-time traffic conditions ▲recent driving distribution, so that the AI logic is built to enable 'fast and maximized matching between passengers and drivers.'"

We explored possible dispatch scenarios through a Q&A.

Q. There is a regular taxi at a closer straight-line distance to the passenger's departure location, but a KakaoT Blue taxi located farther away in the opposite direction was dispatched?

A. It is not simply based on proximity but real-time traffic volume is analyzed. The opposite direction may have relatively less traffic, and the opposite direction vehicle may reach the passenger's location faster by making a U-turn.

Q. There was an empty regular taxi nearby, but the call was assigned to a KakaoT Blue taxi at a different location?

A. Each vehicle has different conditions and situations. The AI comprehensively considers multiple factors to make accurate matches. Driving patterns, preferences, traffic volume, and passenger destinations all differ, so there are almost no taxis with identical conditions.

Q. What does it mean that each taxi has different conditions?

A. For example, a taxi may have a 'selective acceptance history.' If a driver has rejected calls for certain passenger destinations they do not prefer, the AI recognizes that vehicle as having a low acceptance probability.

Q. What if there is no selective acceptance history?

A. It may be a taxi subject to business area restrictions. For example, if the passenger's departure and destination are both in Bundang, Gyeonggi-do, Seoul-area taxis cannot operate there, so those vehicles will not be assigned.

Q. What other scenarios are there?

A. If a driver has not received good ratings from passengers, even if they are within the same ETA (estimated time of arrival) radius, a driver with better ratings is more likely to be assigned. Also, when a passenger calls using an 'automatic payment call' by registering a credit card for fare payment, the call is assigned only to taxis that have completed automatic payment authentication.

This content was produced with the assistance of AI translation services.