June 21 ~ 22, 2025, Sydney, Australia
Omar Hujran, Nuseiba Altarawneh, Statistics and Business Analytics United Arab Emirates University, Al-Ain, UAE
Until recently, scholarly focus on the post-adoption phase, specifically users continued usage of chatbot services, has been notably limited. Recognizing this gap in the existing literature, the current research endeavors to develop an integrative model. This model extends the theoretical framework of the Expectation Confirmation Model (ECM) by incorporating key constructs related to chatbot service quality and anthropomorphism. Both of these concepts are deemed crucial in influencing the continuance usage of chatbot services. As an integral facet of a broader research undertaking, the model will be empirically applied to assess citizens intentions regarding the continued use of chatbot services within context of the United Arab Emirates.
Chatbots, post-adoption, ECM, anthropomorphism, chatbot service quality.
Photis Nanopoulos, National Technical University of Athens, GreeceRetired EUROSTAT, Greece
In this paper we propose the definition of a divergence coefficient between two random variables (Y,X) which provides a general theoretical framework for several classical coefficients used in supervised machine learning algorithms. We use the well-known properties of conditional expectation between two random variables to define a coefficient φ(Y/X) of Y versus X from which we derive as particular cases the classical ID3 ,C4.5 ,CART, alongside other coefficients, proving also the optimality of the solutions they may provide. We also notice the well-known fact that the Bregman divergence is a more general expression of the L2 norm.
Classification methods, optimal partitions, decision trees, Gini coefficient, entropy, Bregman divergences.
Praveen Pilla1 and Gudeangadi Vinaya2, 1Physical Design Engineer, Intel Technology India Pvt. Ltd, Bangalore, 2Soc Design Manager, Intel Technology India Pvt. Ltd, Bangalore
At the SOC level, channels are used to interconnect different IPs together. When the number of wires going through the channels are too many the area of the channels becomes substantial portion of the total area of the SoC. The problem with channels is, they are only carrying the signals. The channels are inefficient way of utilizing the area since channels carry mostly signals with very less utilization. Often to optimize the area, IPs are abutted at the SoC level & channels are replaced by SoC feedthrough wires going across the IPs. While implementing SoCs where IPs are abutted, it is a common practice to run the SoC feedthrough wires across the IPs to reduce the area spent in Channels. In an abutted design methodology, the two IPs connected to each other may not be next to each other due to floorplan constraints. This is the reason SoC feedthrough wires are run across the intermediate IPs. The new age design planning tools can handle creating SoC feedthroughs at IP level. But incase of Multiple Instantiated Blocks (MIBs i.e. multiple instances of the same reference) the design planning tools cannot handle them efficiently & effectively. This paper discusses different scenarios where the tool cannot handle SoC feedthrough creation & proposes solutions to resolve them.
MIB – Multiple Instantiated blocks, SoC: System on Chip, IP: Intellectual Property.
Vinoodhini D1, Ajai Ram2 and Arockia Xavier Annie R3,1, 2 & 3 Anna University, Chennai - 600025, India
Audio verification is a key biometric authentication method used to confirm an individuals identity based on their voice. By tackling issues like fluctuating acoustic conditions, and a range of voice characteristics, this research aims to enhance speaker verification systems. Existing approaches are ineffective in practical situations as security requirements for various applications increase, calling for more reliable solutions. To improve voice feature extraction and analysis, the study makes use of the DF-ResNet architecture, which combines a transformation module with a depth-first search strategy. Speaker verification datasets from actual acoustic settings are used to assess the model. Experimental results demonstrate its effectiveness in improving accuracy while maintaining low computational complexity, making it a viable solution for modern biometric authentication systems.
Biometric Authentication, Depth First Resnet, Transformation Module, Speaker verification.
Ascanio Bernardeschi, Centro Studi Domenico Losurdo, Università Popolare Antonio Gramsci, Roma, Italy
This paper explores the socio-economic implications of Artificial Intelligence, challenging the notion oftechnological neutrality. It critically engages with concepts such as "cognitive capitalism" and thealleged obsolescence of themarxian labor theory of value. AI is shown to intensify labor exploitation byde-skilling, automating intellectual labor, and extending surveillance across global supply chains. WhileAI has the potential to relieve humans from certain cognitive burdens and foster progress in sectors likemedicine or science, its current use under capitalism reproduces class-based inequalities. Is stressed theneed to politically contest capitalist appropriation of AI and reorient it toward collective emancipation.The paper concludes by calling for shorter working hours, transparent and open-source AI development,and the formation of unified political platforms that empower a fragmented working class. The rise of AIshould not lead to techno-pessimism or neo-Luddism, but to a renewed class struggle adapted to thedigital age.
Economics, Finance, Marxian Theory, Open source, Generative AI.