Uncovering the Hidden Cost of Fake News Detection

The world has become overtaken by misinformation, and deep learning gives effective tools. CNN + BiLSTM hybrid networks, in particular when trained with the Harris Hawks Algorithm, have proven to be the best in fake news detection — even with attainable results of up to 98.89 percent.

However, is performance the only truth?

Behind the Measures, the Reality

When it comes to machine learning, the majority of the research talks about measuring such metrics as accuracy, precision, recall, and F1 score, since they are essential to benchmark success.

But there is one metric that people do not always consider: environmental cost.

It is large models that take huge amounts of energy to train.
Indeed, it takes one training run of a deep model to produce the amount of CO₂ emissions that five gasoline-driven cars would use during their lifetime.

And this rarely gets talked about.

Ethical Blind Spot

In aspirations to use AI to do social good, such as fight fake news, the development ought to reflect that as well. Ignoring emissions, electricity, and environmental impact of training a model results in an ethical blind spot in research.

Is the development of one disaster of the world leading to the creation of another?

Toward Responsible AI

Performance is not enough.
It is time to think of:

• ✅ Energy-efficient training

• ✅ Less time with more intelligent algorithms

• ✅ Carbon footprint disclosure

Intelligible AI is not only concerned with precision — but also with being responsible.

Well, what is the big question we should be asking?

When your model says it is 98.89% accurate, you would consider the following question:

What is the price?

This is because we are creating systems to discover the truth —
they ought not to be concealing theirs.

Wish to get more into it?

Join our next webinar:
Fake News Detection Goes Green
📅 https://scholarscolab.com/webinar-on-fake-news-detection-by-optimized-deep-learning/

Also Watch https://www.youtube.com/watch?v=-lHMhOy0bRE