The construction industry is adopting artificial intelligence (AI) rapidly and it is easy to see why –the technology is able to take data from the past and make help to predict the future. Here, it makes it easy for people to think about safer and better workplaces. Even with this impressive advancement, technology adoption is coming with a number of ethical dilemmas.
How effectively can an organization guarantee utmost data privacy and governance? What is the right transparency level for different categories of data? How can you guarantee fairness and discrimination-free work sites? These are only a few of concerns, which have made people call for stringent rules in the application of AI in different areas.
What is the Way Forward? The Ethical Guidelines for AI Application
While there is no doubt that AI is a powerful technology, the importance of proper regulation cannot be overemphasized. Ethical interventions are needed to prevent its misuse, especially customer data. So, what initiatives are in place for safeguarding AI applications? One of them is the EU Ethics Guidelines for Trustworthy Artificial Intelligence. According to these guidelines, AI needs to be:
- Lawful and adhering to all applicable regulations and laws.
- Respecting ethical principles and values.
- Robust from both technical and social perspectives.
Factoring these needs, the guidelines list seven crucial requirements that every AI system that is applied within the EU jurisdiction should meet in order to be labeled trustworthy.
- Human agency oversight: The design of any AI system should be aimed at empowering people to help them make the right decisions and advancing their rights. Furthermore, it should promote better oversight that can be achieved via direct human interaction and ongoing human control.
- Technical robustness and safety: Every AI system should be secure and resilient. Particularly, it should have a good backup plan to help reset the system in case something goes wrong. In addition, it should be reliable, accurate, and reproducible.
- Privacy and data governance: In addition to respect for privacy and data protection, the AI system should have a good governance system factoring in quality and integrity of the data.
- Transparency: The data system that any AI uses should have robust traceability mechanisms for enhanced transparency. Again, the mechanisms used in AI should be explained in a way that stakeholders can understand. Particularly, it should be easy to comprehend the capabilities and limitations.
- Diversity, and non-discrimination, and fairness: This requirement is aimed at ensuring the AI system does not promote bias, especially against the vulnerable groups. Therefore, any AI system should be available to all, irrespective of their disability.
- Social and environmental well-being: While the focus of the AI systems is to address issues that have been identified in the past and present, it should also be targeted at helping the future generations. Therefore, they should be as sustainable as possible. This means they must factor other living organisms, with the target being to reduce possible negative impacts.
- Accountability: Any AI system should have a mechanism for promoting responsibility and accountability. For example, it should be auditable to help with the assessment of data, algorithms, and data processes.
These guidelines are likely to pose some challenges for construction businesses in the near future. For example, there is a possibility of breaches of privacy, transparency and non-discriminatory requirements outlined on the guideline. Also, the autonomous paradigm shift in the construction industry might come with some conflicts as the sector crafts its own ethical and operating standards.
Although the application of AI in the construction industry is expected to continue defining the sector because of its huge potential, the emerging regulations are likely to result in major ethical dilemmas. How effectively these dilemmas are addressed will define the pace at which AI can get adopted in the construction industry.