Key Points
- AI experts say Uber's algorithms could easily have prevented the dramatic price surges experienced by Sydney commuters
- They say it's clear Uber's algorithms are designed to boost company profits.
- Uber says it immediately lowered and capped the surge as soon as it was aware of the issue.
Artificial intelligence experts have slammed Uber's algorithms that allowed the company to charge stranded Sydneysiders exorbitant fares after Wednesday's train shutdown, claiming simple tweaks in code could easily have prevented the crisis.
Two of Australia's prominent voices in AI say it's clear Uber's algorithms place company profits over customer satisfaction. Instead they should have enabled fast intervention in the event of dramatic price increases.
A shutdown in the Sydney trains network attributed to a communication issue left commuters scrambling for ride-sharing services, with some people saying they were quoted prices of more than $300 for a trip of around half an hour, or more than $200 for a fare usually priced at $20.
Surge pricing is a dynamic pricing method where prices are temporarily increased as a reaction to increased demand and mostly limited supply.
It's a common practice among ride-share companies as well as hotels and airlines and is legal under Australian Consumer Law, provided businesses are clear about the price consumers will pay and can provide reason for the price hikes.
On its website, Uber claims the practice incentivises more drivers to work in times of high demand, which in turn reduces rates and ensures reliability.
Uber on Thursday released a statement blaming Transport NSW for the lack of notification regarding the train shutdown and stating it lowered the surge as soon as it became aware of the situation. It also promised to refund customers who were overcharged.
"In the past we have been alerted by Transport for NSW when there were Sydney-wide transport issues, however in this instance we were not informed of the complete outage on the NSW train network until well after it began, " the statement said.
"As soon as our team became aware of the train disruption, we immediately lowered and capped surge to still incentivise driver-partners who were helping Sydneysiders get home, while making rides more affordable for those stranded."
Uber adds that it has measures in place that stop the surge algorithm from reaching a certain level, and says some of the price hikes reported were "impossible" under these measures. It also claims to let riders know about the price surge so they have the option to cancel.
But AI experts say while it's easy to blame price surging incidences on algorithms, the same AI can also prevent them happening in the first place.
Price surging or price gouging?
It's hard to say when price surging veers into price gouging, particularly when AI algorithms aren't open source.
"It's not clear whether the price is literally the right amount to get enough drives out to service the domain, or whether it's price gouging. We just don't know. Obviously, if it's price gouging, it's problematic," deputy director of the UTS Data Science Institute Adam Berry told SBS News.
He says it would have been easy for Uber to see quickly that prices were rising by a certain amount. And a simple algorithm could have fixed a price cap or alerted a human to intervene.
"Uber has predictive models that can understand typical behaviours at different times of day in different areas. Certainly it would have the capacity for those algorithms to flag atypical behaviours."
"It's one line of software code to say, the price is 10 times higher than it usually is and send an alert. One line of code can fix this forever. But obviously Uber has chosen not to do that.
"Some businesses deliberately make AI this mystical, otherworldly thing when in fact, it is pretty straightforward."
A shutdown in the Sydney trains network left commuters scrambling for Uber rides, with claims of skyrocketing fares. Source: AAP / Joel Carrett
"AI doesn't take human experiences or emotions into account, which is unfortunate. But companies can absolutely build algorithms around customer satisfaction and fairness, and they can use human oversight to ensure that their pricing practices are fair and ethical," he told SBS News.
"We need to use human interaction and human judgement and not make a decision just based on data."
Calls to make algorithms open source
Mr Berry and Mr Mirjalili agree the wider problem is a lack of transparency when it comes to algorithms; we simply don't know how companies' codes and priorities are set up.
"Businesses are very reluctant to reveal anything about how those algorithm's work so it's very difficult for anyone such as policymakers, users of the service or regulators to comment on whether what they're doing is ethical or not, because we just don't know," Mr Berry said.
Ideally, the creation of company algorithms should involve community or customer consultation, he added.
"Certainly, there is a trend that these algorithms are built very quickly, often with a profit mindset behind them. If you're moving at speed, you typically don't have six months of community consultation. Because then you'll found 45 businesses have released their tech without doing that."
Mr Mirjalili says he hopes there will one day be recommendations for companies to make their algorithms open source.
"This will allow us to see when companies are truly prioritising profit over the quality of the service," he said.
