AI Bias: How Can Blockchains Promote Equity?

Projects based on artificial intelligence (AI) are quickly becoming a fundamental component of the contemporary technological landscape, assisting in decision-making across a range of industries, including finance and healthcare. Nevertheless, despite notable advancements, AI systems have their shortcomings. One of the most pressing challenges confronting AI today is the issue of data biases, which pertains to the existence of systematic inaccuracies within a dataset that can lead to distorted outcomes when training machine learning models.

Given that AI systems are heavily dependent on data, the quality of the input data is critically important, as any form of biased information can introduce prejudice into the system. This can further reinforce discrimination and inequality within society. Thus, maintaining the integrity and impartiality of data is vital.

For instance, a recent article discusses how AI-generated images, particularly those produced from datasets predominantly influenced by American sources, can misrepresent and standardize the cultural context of facial expressions. It provides several examples of soldiers or warriors from different historical eras, all displaying the same American-style smile.

An AI generated image of Native Americans. Source: Medium

Furthermore, the pervasive bias not only fails to reflect the diversity and subtleties of human expression but also poses a risk of erasing essential cultural histories and meanings, potentially impacting global mental health, well-being, and the richness of human experiences. To counteract such biases, it is crucial to integrate diverse and representative datasets into AI training methodologies.

Several elements contribute to biased data in AI systems. Firstly, the data collection process may be flawed, resulting in samples that do not accurately represent the target population. This can lead to the underrepresentation or overrepresentation of specific groups. Secondly, historical biases can infiltrate training data, perpetuating existing societal prejudices. For example, AI systems trained on biased historical data may continue to reinforce gender or racial stereotypes.

Lastly, human biases can unintentionally be introduced during the data labeling phase, as labelers may possess unconscious biases. The selection of features or variables utilized in AI models can lead to biased results, as some features may correlate more strongly with certain groups, resulting in unfair treatment. To address these challenges, researchers and practitioners must recognize potential sources of skewed objectivity and actively strive to eliminate them.

Can blockchain make unbiased AI possible?

While blockchain technology can assist with certain aspects of maintaining neutrality in AI systems, it is not a comprehensive solution for completely eradicating biases. AI systems, such as machine learning models, can develop discriminatory tendencies based on the data they are trained on. Moreover, if the training data contains various predispositions, the system is likely to learn and replicate them in its outputs.

Nonetheless, blockchain technology can play a role in addressing AI biases in its own distinct ways. For example, it can help ensure data provenance and transparency. Decentralized systems can trace the origin of the data used to train AI systems, promoting transparency in the information collection and aggregation process. This can assist stakeholders in identifying potential sources of bias and addressing them.

Recent: Why join a blockchain gaming guild? Fun, profit and create better games

In a similar vein, blockchains can enable secure and efficient data sharing among multiple parties, facilitating the creation of more diverse and representative datasets.

Additionally, by decentralizing the training process, blockchain can allow multiple parties to contribute their own information and expertise, which can help reduce the influence of any single biased perspective.

Maintaining objective neutrality necessitates careful attention to the various phases of AI development, including data collection, model training, and evaluation. Furthermore, ongoing monitoring and updating of AI systems are essential to address potential biases that may emerge over time.

To gain further insight into whether blockchain technology can render AI systems entirely neutral, Cointelegraph contacted Ben Goertzel, founder and CEO of SingularityNET — a project that merges artificial intelligence and blockchain.

In his perspective, the notion of “complete objectivity” is not particularly useful in the context of finite intelligence systems analyzing finite datasets.

“What blockchain and Web3 systems can provide is not complete objectivity or absence of bias but rather transparency so that users can clearly see what bias an AI system possesses. It also offers open configurability so that a user community can adjust an AI model to reflect the type of bias it prefers and transparently observe what kind of bias it is exhibiting,” he stated.

He further noted that in the realm of AI research, “bias” is not a negative term. Instead, it simply indicates the orientation of an AI system searching for specific patterns in data. However, Goertzel acknowledged that opaque biases imposed by centralized organizations on users who are unaware of them — yet are influenced by them — are a concern that people should be cautious about. He remarked:

“Most popular AI algorithms, such as ChatGPT, are lacking in terms of transparency and disclosure of their own biases. Therefore, part of what is necessary to effectively address the AI-bias issue is decentralized participatory networks and open models that are not just open-source but also include open-weight matrices that are trained, adapted models with open content.”

Similarly, Dan Peterson, chief operating officer for Tenet — an AI-focused blockchain network — informed Cointelegraph that quantifying neutrality is challenging and that certain AI metrics cannot be unbiased because there is no measurable threshold for when a dataset loses its neutrality. In his opinion, it ultimately comes down to the perspective of where the engineer draws the line, and that line can differ from person to person.

“The idea of anything being truly ‘unbiased’ has historically been a difficult hurdle to overcome. Although absolute truth in any dataset fed into generative AI systems may be elusive, we can leverage the tools made more accessible through blockchain and Web3 technology,” he stated.

Peterson mentioned that techniques based on distributed systems, verifiability, and even social proofing can assist in creating AI systems that come “as close to” absolute truth. “However, it is not yet a turnkey solution; these emerging technologies help us advance rapidly as we continue to develop the systems of the future,” he added.

Looking toward an AI-driven future

Scalability remains a major concern for blockchain technology. As the number of users and transactions grows, it may restrict the capacity of blockchain solutions to manage the vast amounts of data generated and processed by AI systems. Additionally, the adoption and integration of blockchain-based solutions into existing AIs present significant challenges.

Recent: Crypto in Europe: Economist breaks down MiCA and future of stablecoins

First, there is a lack of understanding and expertise in both AI and blockchain technologies, which may impede the effective development and deployment of solutions that integrate both paradigms. Second, persuading stakeholders of the advantages of blockchain platforms, particularly regarding unbiased AI data transmission, may be difficult, at least initially.

Despite these obstacles, blockchain technology possesses substantial potential for leveling the rapidly evolving AI landscape. By utilizing key attributes of blockchain — such as decentralization, transparency, and immutability — it is feasible to diminish biases in data collection, management, and labeling, ultimately leading to more equitable AI systems. Consequently, it will be intriguing to observe how the future unfolds from this point onward.