What is “data altruism” and will it make AI better?
Definition
In a bid to increase the amount of data available in Europe to compete with Asia and America, the European Commission introduced the idea of “data altruism”.
"With the ever-growing role of industrial data in our economy, Europe needs an open yet sovereign single market for data,".
Sharing data
Europe’s proposed Data Governance Act will allow business and research organisations to tap the potential of such vast data volumes.
Data altruism encourages individuals and companies to share data for the common good, on the assumption that more data is better data. Whilst there is some truth in the old saying that “there’s no data like more data”, it doesn’t make AI applications perfect. At the end of the day they are simply statistical pattern matching algorithms and they have no intelligence nor reasoning abilities.
Limits of AI
"Scientists may train AIs to detect hate speech on data where such speech is unusually high, such as white supremacist forums. However, when this software is exposed to the real world, it can fail to recognize that black and gay people may respectively use the words "black" and "gay" more often than other groups. "Even if a post is quoting a news article mentioning Jewish or black or gay people without any particular sentiment, it might be misclassified as hate speech," Ren says. In contrast, "humans reading through a whole sentence can recognize when an adjective is used in a hateful context.
Vested interests
To avoid the vested interests of BigTech, the EU wants intermediaries to make available data sharing across member states and sectors without benefit to themselves, to increase public trust. Public sector bodies would be encouraged to use proposed mechanisms with appropriate data protection to share what is normally sensitive data for both commercial and non-commercial purposes.
Private data too
Although under the EU rules, consent is required, the danger is that individuals may feel pressured into making available their private data. Acting outside of these proposed new rules, the UK’s National Health Service has simply required people to opt out of the pooling and sharing of their personal health data. Most citizens were unaware of the deadline for this and this assumed consent to “data altruism” calls into question the rights to privacy of personal data and the idea that a human being ought to have sovereignty over data that derives from them.
Public benefit
Whilst data altruism might lead to public benefit through better planning, health diagnostics or disease prevention, it is not a given. AI based applications do fail and don’t always help where they were expected to, as has been reported in their use during the Covid Pandemic. Image recognition systems used in both facial recognition and medical diagnostics can be very brittle, misclassifying an image that it hasn’t seen before or that has been manipulated.
More data will not necessarily create more trustworthy AI yet the illusion that it can, could run the risk of encouraging more and more people to rely on it, to the detriment of humanity, especially when it fails, as we illustrated above.
References
https://spectrum.ieee.org/ai-failures