Responsible AI Licence (RAIL) for AMIVA-F

Preamble:

This Licence governs the use, replication, and modification of AMIVA-F, an AI tool designed for classifying single-point variants in human FlnC. Recognizing the potential of AMIVA-F and derivatives, while acknowledging its ethical and clinical implications, this Licence aims to ensure responsible and equitable use.

1. Ethical Risks and Concerns:

1.1 Misinterpretation of Data: The risk of misinterpreting classifications can lead to erroneous scientific conclusions and/or medical and clinical decisions.

1.2 Bias and Fairness: The inherent biases in limited available training data that arise based on validation of population frequency based cohorts from databases like Clinvar and gnomAD can lead to skewed or unfair predictions across different genetic populations due to unbalanced representation.

1.3 Privacy and Confidentiality: The potential misuse of sensitive genetic information, leading to privacy violations.

1.4 Dependence: Over-reliance on AMIVA-F predictions may erode traditional scientific, as well as medical, expertise and critical analysis skills.

2. Usage Restrictions:

2.1 Authorized Users:

Only individuals and organizations that agree to adhere to this RAIL are authorized to use AMIVA-F.

Users must possess the requisite background in genetics, medicine or a related field to ensure informed interpretation of the tool's outputs.

2.2 Replication and Modification:

Replication and modification of AMIVA-F for potential derivatives are permitted solely for non-commercial, non-clinical, research, and educational purposes.

Any modifications must be clearly labeled and must contain the README.md as well as the LICENSE file easily accessible and available.

2.3 Prohibited Uses:

AMIVA-F shall not be used for any purposes that discriminate against individuals or groups based on genetic characteristics.

The tool shall not be used as the basis for medical or clinical decisions, neither with, nor without human oversight

3. FAIR and FAT Principles:

3.1 Fairness:

Efforts must be made to ensure the tool's predictions are applicable across diverse genetic backgrounds. While intrinsically biased due to the underlying patient cohorts used in training, effort shall be made for future datasets to be more balanced towards an equal sex/race proportion.

3.2 Accountability:

Users must maintain records of AMIVA-F's application in research and shall not rely on it in clinical settings.

3.3 Transparency:

Publications or reports derived from the use of AMIVA-F must disclose the tool's involvement and any limitations, e.g its intrinsical bias based on the utilized databases and their genetic populations for training.

4. Compliance and Enforcement:

4.1 Users must ensure compliance with this Licence's terms.

4.2 Violations of this Licence may result in revocation of the right to use AMIVA-F and potential legal action.

5. Liability and Disclaimer:

5.1 The creators of AMIVA-F disclaim any liability for damages resulting from the tool's use, including but not limited to, errors in prediction or misinterpretation of data.

5.2 This Licence does not guarantee the tool's accuracy or suitability for a particular purpose and might deviate depending on sex and/or race.