Computers can identify people using edge detecting algorithms and natural language processing helps speech recognition, transcription and translation with linguistics.
This is different to traditional logical problem solving approach. AI programs the machine to learn itself with the dataset we input to derive the most likely logic to predict result for a given input with some probable error rate. Unlike a logical program, the result is not 100%. Even for the same data set result can vary.
In science, technology, medicine and business AI offers enormous advances as well questions in ethics.
When we are Training a machine to predict which convicted felons will reoffend, the criminal data we input are biased against low-income people. Therefore the result could be biased against low-income people.
Making websites more addictive with AI biased content delivery is biased towards revenue but unfair to users. A research done on Facebook using six identical profiles subscribing to identical news sources found that differently curated content on their news feeds.
AI can be used to write fake reviews on products or fake comments on social media posts that could help go viral, making it harder for the truth to get out.
Analysis of varying sentiments towards an election, AI could predit countering, faking , biasing content for campaigns of richer condidates giving unfair advantages.
Machines become better at achieving the goal they learned to pursue in their training environment, but it may not the outcome we expect in actually setting elsewhere.
Unlike logical systems AI systems are developped with least knowledge of the problem and logic of the
solution. It means, the behavior of an AI program may not be what we always anticipate.
AI presents newer threats for human rights in privacy, security, freedom of expression, freedom of association, non-discrimination, right to work and access to public services.
Therefore we need human rights-based approach to AI as there are no established methodologies to asses their impact on human rights.
AI need to identify who is being potential for discrimination. The solution would be human-centric AI.
The European Commission
EU strategy on AI (April 2018), makes specific reference to the need to invest in human-centric, inclusive approach to AI.
The first draft of AI ethics guidelines to the Commission by EU high level expert group, address values protected by the Charter of Fundamental Rights, such as privacy and personal data protection, human dignity, non-discrimination and consumer protection.
The guidelines ask all stakeholders to evaluate possible effects of AI on human beings and the common good, further ask ensuring human centric AI development, deployement and ethical use grounded on fundamental human rights, social values and and justice.
This requires stakeholders to draw up ethical guidelines with ethics-by-design for embedding transparency and explainability into AI programs based on human rights, ethics and justice.
Niranjan Meegammana
It
Niranjan Meegammana
HR Researcher & Technologist
Digital Human Rights
No comments:
Post a Comment