Artificial intelligence (AI) is undoubtedly one of the most significant technological advancements of this era. It has revolutionised the way to interact with technology, influencing medicine to finance, entertainment to education.
Despite the ubiquity of AI, heightened by the emergence of generative AIs, biases on the technology remain a major challenge. It is imperative that education, in all its dimensions, tackles this issue proactively and in an informed manner.
The role of education
Education plays a crucial role in raising awareness about AI biases. Educational institutions, from primary schools to higher education institutions, can integrate AI and ethics modules into their curricula. This allows students to gain a deep understanding of the implications of AI biases and their impact on society. By developing this understanding, students are better prepared to tackle ethical challenges as AI becomes ubiquitous.
Furthermore, education can promote a culture of responsibility in AI development. Future AI developers and users must be aware of the potential consequences of their actions. Learning AI ethics should be a fundamental component for those entering the fields of Computer Science and Data Science. They must understand that they have a critical role to play in preventing AI biases from becoming a bigger problem.
Encouraging teachers to teach AI ethics
For education to be truly effective in combating AI biases, teachers must be supported and trained to teach AI ethics to students of all ages. Students should be encouraged to question how AI affects their daily lives and how they can contribute to mitigating biases.
Teachers can play an essential role in guiding students to think critically about AI and helping them develop critical skills to evaluate automated systems. For this, they can practise with responses provided by generative AIs and challenge their accuracy and relevance.
Acting without delay
Education plays a key role in the fight against AI biases. The visible or hidden ubiquity of AI in society means that the decisions it makes and the recommendations it provides have a substantial impact. If steps are not taken to mitigate biases, societies risk perpetuating inequalities and injustices, whether consciously or unconsciously.
Societies cannot afford to wait to act, as the beneficial effects of education on the topic may take time to manifest. Education must address this complex issue and ensure that AI is used ethically and equitably for a more responsible society.
Understanding AI biases
Biases in AI originate during the model training stage. If the training data contains prejudices or discriminations, AI can replicate them indiscriminately. For instance, facial recognition systems have been accused of exhibiting higher error rates for people of colour, partly due to biased training datasets. This underscores the crucial importance of addressing this issue.
To understand AI biases, it’s essential to grasp how they emerge. AI models learn from real data which they use to make decisions or recommendations. If this data is tainted with stereotypes, injustices or discriminations, AI can perpetuate these problems. Therefore, it is crucial to address biases at the source, which is the training data.
An instructive project
The ‘Gender Shades’ project assesses the accuracy of AI-powered gender classification software. This study revealed that companies subjected to these evaluations achieved significantly better facial recognition results for men than for women, with error rate differences ranging from 8.1 percent to 20.6 percent.
A study by the Massachusetts Institute of Technology (MIT) examining facial analysis software also revealed a surprising error rate of 0.8 percent for light-skinned men and 34.7 percent for dark-skinned women.
The project’s findings highlighted the potential for technology misuse, whether by authoritarian governments, malicious individuals, or opportunistic companies. Ultimately, the project emphasised the vulnerability of algorithms to corruption and bias.
Promoting diversity in AI
Another critical aspect of combating AI biases is promoting diversity. Currently, there is an evident imbalance in terms of gender, race and socio-economic backgrounds among AI developers. This homogeneity can contribute to unintentional biases in AI systems.
Education should take a leading role in encouraging diversity in the AI field. Educational institutions, in collaboration with businesses and organisations, can take steps to recruit more underrepresented individuals into AI careers. Mentorship programmes, scholarships and outreach initiatives can help attract a diverse range of talent to AI careers.
Training AI developers to design equitable systems
Business schools and engineering schools have an essential role in introducing common curricula that emphasise AI ethics and bias reduction. Computer Science students must learn to design fair AI systems and assess training data for potential biases. They should be able to develop AI models that not only maximise prediction accuracy but also consider ethics.
The writer is the chief digital officer, Excelia Group.
