You’re involved in a project called ClimateBert, which uses AI to assess
climate-related corporate disclosures’ increasing transparency. How does that work?
ClimateBert is the product of joint research with some great researchers at ETH and the
University of Nürnberg. What we do is take an algorithm that was developed by Google in
2018—which is basically a neural network that can understand text in its context—and we
further develop and train this algorithm to understand climate-related text. So basically we’re
feeding something like 1.4 million paragraphs related to climate, and by training this network
we are able to ask very interesting questions regarding the disclosure of climate-related risks
You’ve been examining over 800 companies that support the Task Force on
Climate-Related Financial Disclosures, also known as TCFD, and you concluded that
what they’re doing was mostly cheap talk.
Yes, basically we wrote two articles using ClimateBert, and we analyzed annual reports,
because I think that’s a more important medium because these are legal documents
compared to TCFD reports, where companies can basically write a little bit of whatever they
want. So we looked at whether companies are stating commitments in terms of climate
mitigation and climate adaptation measures that they’re taking and whether these
commitments are actually specific, because you can write some boiled up language,
anything about what you will do in the future. But we really wanted to analyze specific
commitments, and then we related these specific commitments to the non-specific
commitments. We created something that I call CTI, the Cheap Talk Index, and we saw that
over the years, and especially after the Paris Agreement, the Cheap Talk Index has been on
the rise. So companies tend to write more about climate but not necessarily in a very
informative way so investors can really grasp the exposure of the companies to climate risk.
So can you give us an example when you would say “this is actually greenwashing”?
Greenwashing is kind of a difficult concept. In my research—not necessarily related to
finance—I did some work on fact-checking. I developed algorithms on fact-checking, and
there it is slightly easier because there’s a black-and-white decision that we make on a
claim. But in greenwashing, there are many shades of green, and now there’s FINMA, which
released a couple of examples of what they consider greenwashing. And it’s probably good
to, as time goes by, get a better feeling of what greenwashing really is and this is actually the
next goal of this research project with ClimateBert—that we develop algorithms that deal
with greenwashing detection.
And coming back to TCFD disclosures: How could these be improved?
Don’t get me wrong. I’m basically a capitalist by tooth and claw, but when markets fail what
is actually needed is government intervention. And from my experience and from my work as
an academic I conclude that what is needed is that we need to transfer these TCFD
reporting requirements into a regulatory framework—and that’s still not enough. I think that
the content of the TCFD reporting needs to put more focus on the reporting of metrics and
targets because in the end this is what counts for our climate and for the transition into a
more sustainable economy.
So you ask for the government to step in, for regulators to establish standards. What
is being done in this regard?
I think that there are some countries that now support TCFD that they will put that into some
regulatory framework, so that’s definitely applaudable. But then there needs to be more
ambitious goals that need to be formulated: harder targets in terms of emission control. And
when you look at what is said in the Paris Agreement in article 2.1c about where climate
finance can contribute to the transition—I think [in] most countries there’s a big gap between
fulfilling these requirements, and this also holds for Switzerland. So much more needs to be
And how are you helping with your projects? I don’t know if you’re collaborating with
Currently, I’m not involved with the government. But what we do is try to provide instruments,
also for regulators, by, for instance, ClimateBert. This is an algorithm that is downloadable at
no cost and can be further trained to ask important questions that are needed in order to
provide transparency into the sustainable-finance world.
Switzerland is aiming to become a center for sustainable finance. Why would this be
the ideal place for it?
I think there are three parties involved: the government, the financial industry, and academia.
And sustainable finance consists of two words: sustainability and finance. Switzerland has a
long tradition in finance, so therefore I think the prerequisite for becoming a leader in
sustainable finance is all there. The academic world has a very strong reputation in terms of
finance through the Swiss Finance Institute, and there are also different centers across the
country. For instance, here at the University of Zurich we are very strong in sustainable
finance, which was pioneered by Marc Chesney. So we are in an optimal position to interact
with the industry on sustainable finance issues, and the finance industry obviously also has
put a lot of emphasis on sustainability. However, I think that we still need to be a little bit
more ambitious in defining our goals for the future and to ensure sustainable development.