1
00:00:00,050 --> 00:00:00,710
Case study.

2
00:00:00,710 --> 00:00:06,020
Global AI Auditing framework ensuring ethical and accountable deployment at Technova.

3
00:00:06,050 --> 00:00:11,240
The necessity for a global artificial intelligence auditing framework propels us into understanding

4
00:00:11,240 --> 00:00:17,480
not just the technological, but also the ethical, legal and social dimensions of AI integration in

5
00:00:17,480 --> 00:00:18,530
today's world.

6
00:00:19,190 --> 00:00:24,200
Picture an international conglomerate, Technova, which operates in industries ranging from healthcare

7
00:00:24,200 --> 00:00:28,010
to finance, all powered by advanced AI systems.

8
00:00:28,520 --> 00:00:34,880
As Technova expands, the complexity of managing and auditing these AI systems grows, raising critical

9
00:00:34,880 --> 00:00:37,850
questions about their ethical and accountable deployment.

10
00:00:39,380 --> 00:00:44,870
In Technova headquarters, Maria, the Chief Ethics officer, and Raj, the head of AI development,

11
00:00:44,870 --> 00:00:50,450
confront an emerging issue a bias discovered in an AI system used for hiring.

12
00:00:50,960 --> 00:00:56,510
This system, intended to streamline recruitment across multiple countries, appears to be systematically

13
00:00:56,510 --> 00:00:59,360
favouring certain demographic groups over others.

14
00:00:59,810 --> 00:01:05,860
Maria and Raj are faced with the task of auditing the AI system to ensure compliance with ethical standards

15
00:01:05,860 --> 00:01:07,360
and legal requirements.

16
00:01:07,780 --> 00:01:13,690
Here, the conversation begins with the fundamental question how can technology ensure transparency

17
00:01:13,690 --> 00:01:18,370
in the functionality and decision making processes of its AI systems?

18
00:01:19,330 --> 00:01:25,870
Transparency entails the clear documentation and disclosure of how the AI operates, including the data

19
00:01:25,870 --> 00:01:28,270
sources and potential biases present.

20
00:01:28,810 --> 00:01:35,200
Maria suggests creating a detailed log of the AI's decision making process, which Raj supports by proposing

21
00:01:35,200 --> 00:01:40,750
the inclusion of explainable AI techniques that can outline how decisions are reached.

22
00:01:41,140 --> 00:01:45,700
This approach provides stakeholders a window into the algorithm's inner workings.

23
00:01:45,730 --> 00:01:48,820
However, achieving full transparency is challenging.

24
00:01:48,820 --> 00:01:54,580
What specific protocols can Technova adopt to ensure comprehensive documentation of their AI systems

25
00:01:54,580 --> 00:01:57,130
without compromising sensitive information?

26
00:01:58,750 --> 00:02:03,670
Raj mentions leveraging standards developed by organizations like the International Organization for

27
00:02:03,670 --> 00:02:10,120
standardization and the Institute of Electrical and Electronics Engineers to guide their auditing procedures.

28
00:02:10,660 --> 00:02:16,330
These standards offer a structured foundation for evaluating AI systems, focusing on performance,

29
00:02:16,360 --> 00:02:18,460
safety, and risk management.

30
00:02:18,850 --> 00:02:25,780
For instance, ISO, IEC, JTC one, SC 42 guidelines could be instrumental in shaping Tennovas auditing

31
00:02:25,780 --> 00:02:26,680
protocols.

32
00:02:26,950 --> 00:02:33,190
But how can Technova ensure these standards are effectively integrated into their existing processes

33
00:02:33,190 --> 00:02:35,290
across various jurisdictions?

34
00:02:36,730 --> 00:02:42,640
Maria emphasizes the importance of establishing an independent auditing body to ensure an unbiased assessment

35
00:02:42,640 --> 00:02:44,050
of the AI systems.

36
00:02:44,560 --> 00:02:50,710
This body, comprising experts in AI ethics, law, data science, and cybersecurity, would conduct

37
00:02:50,710 --> 00:02:56,470
rigorous evaluations and have unfettered access to the AI's code, training data, and decision making

38
00:02:56,470 --> 00:02:57,430
algorithms.

39
00:02:58,030 --> 00:03:02,890
Raj suggests that this body should also be empowered to recommend corrective actions.

40
00:03:02,920 --> 00:03:08,440
How can Technova balance the independence of this auditing body with its need for internal oversight

41
00:03:08,460 --> 00:03:09,450
and control.

42
00:03:11,040 --> 00:03:16,590
A significant hurdle is the variation in regulatory landscapes across different countries where Technova

43
00:03:16,620 --> 00:03:17,460
operates.

44
00:03:17,850 --> 00:03:23,730
While stringent data protection laws in Europe necessitate rigorous compliance checks, other regions

45
00:03:23,730 --> 00:03:25,740
have more lenient regulations.

46
00:03:26,100 --> 00:03:32,400
Maria outlines the need for international cooperation to harmonize regulations, ensuring a unified

47
00:03:32,400 --> 00:03:35,340
framework that Technova can adopt globally.

48
00:03:35,370 --> 00:03:41,220
What strategies can Technova employ to foster international collaboration and ensure consistent ethical

49
00:03:41,220 --> 00:03:43,500
standards across its operations?

50
00:03:44,940 --> 00:03:48,810
Pre-deployment and post-deployment audits are central to this framework.

51
00:03:49,170 --> 00:03:54,180
Pre-deployment audits would scrutinize the EHS design and development phases, identifying potential

52
00:03:54,180 --> 00:03:55,710
risks and biases.

53
00:03:55,710 --> 00:04:02,430
Raj advocates for a robust testing phase where the AI is evaluated under varying conditions to ensure

54
00:04:02,460 --> 00:04:04,170
fairness and compliance.

55
00:04:04,500 --> 00:04:10,710
Post-deployment audits, Maria adds, would monitor the AI's performance over time, identifying any

56
00:04:10,740 --> 00:04:11,880
emerging issues?

57
00:04:11,880 --> 00:04:17,700
What mechanisms can Technova implement to ensure these audits are thorough and continuous, addressing

58
00:04:17,700 --> 00:04:20,130
both immediate and long term risks.

59
00:04:21,330 --> 00:04:25,200
The socio ethical implications of AI are another critical aspect.

60
00:04:25,770 --> 00:04:31,260
The hiring bias issue at Technova highlights the need for auditors to consider the broader societal

61
00:04:31,260 --> 00:04:33,120
impact of AI decisions.

62
00:04:33,750 --> 00:04:39,240
The auditing body should assess whether the AI respects human rights and promotes social equity.

63
00:04:39,720 --> 00:04:46,200
Raj proposes a series of impact assessments that explore the AI's effects on employment and social dynamics.

64
00:04:46,200 --> 00:04:52,230
How can Technova ensure these socio ethical evaluations are integrated into the auditing process without

65
00:04:52,230 --> 00:04:54,270
hindering technological innovation?

66
00:04:56,220 --> 00:05:02,550
Stakeholder engagement and public participation play a pivotal role in enhancing transparency and trust

67
00:05:03,330 --> 00:05:08,640
by involving affected communities and industry representatives, Technova can gain valuable insights

68
00:05:08,640 --> 00:05:11,580
into the real world impact of its AI systems.

69
00:05:11,970 --> 00:05:18,230
Maria suggests organizing regular public consultations and focus groups to gather diverse perspectives.

70
00:05:18,620 --> 00:05:24,050
How can Technova implement a stakeholder engagement strategy that is both inclusive and effective in

71
00:05:24,050 --> 00:05:26,090
driving ethical AI use?

72
00:05:27,320 --> 00:05:33,200
Investing in education and capacity building initiatives is essential for sustaining a robust auditing

73
00:05:33,200 --> 00:05:33,980
framework.

74
00:05:34,520 --> 00:05:40,820
Technova can develop training programs for auditors, developers, and regulators, focusing on AI ethics

75
00:05:40,820 --> 00:05:42,590
and regulatory compliance.

76
00:05:43,040 --> 00:05:48,830
Raj mentions the AI Governance Professional certification as a potential avenue for structured learning.

77
00:05:48,830 --> 00:05:54,050
How can Technova ensure continuous professional development and keep its workforce updated with the

78
00:05:54,050 --> 00:05:57,050
latest advancements in AI ethics and governance?

79
00:05:59,150 --> 00:06:03,980
To support these initiatives, advanced auditing tools and technologies are necessary.

80
00:06:04,670 --> 00:06:10,730
Automated tools can facilitate large scale evaluations, identifying patterns of bias and assessing

81
00:06:10,760 --> 00:06:12,200
algorithmic fairness.

82
00:06:12,770 --> 00:06:17,750
Raj highlights the potential of blockchain technology in And enhancing the traceability of the auditing

83
00:06:17,750 --> 00:06:18,650
process.

84
00:06:19,250 --> 00:06:24,830
How can Technova integrate these advanced tools into their auditing framework to improve efficiency

85
00:06:24,830 --> 00:06:26,060
and reliability?

86
00:06:27,290 --> 00:06:33,560
In conclusion, as Maria and Raj navigate the complexities of building a global AI auditing framework

87
00:06:33,560 --> 00:06:39,530
at Technova, they emphasize the importance of transparency, fairness, and accountability.

88
00:06:40,550 --> 00:06:46,280
By adopting international standards, establishing independent auditing bodies, fostering international

89
00:06:46,280 --> 00:06:52,220
collaboration and engaging stakeholders, Technova can develop a comprehensive framework that addresses

90
00:06:52,220 --> 00:06:55,430
the ethical and accountability challenges posed by AI.

91
00:06:55,460 --> 00:07:01,010
Investing in education and advanced auditing tools further supports the effective implementation of

92
00:07:01,010 --> 00:07:07,340
this framework, ensuring that Tennovas AI systems operate ethically and responsibly across different

93
00:07:07,340 --> 00:07:08,810
sectors and regions.

94
00:07:09,110 --> 00:07:15,260
This case exemplifies the intricate balance required to foster public trust in AI technologies, while

95
00:07:15,260 --> 00:07:18,050
promoting their responsible and beneficial use.
