class: center, middle, inverse, title-slide .title[ # Artificial intelligence and<br>causal inference ] .subtitle[ ## XLVII Convegno AIE ] .author[ ### Claus Thorn Ekstrøm<br>UCPH Biostatistics ] .date[ ### April 21, 2023<br><svg aria-hidden="true" role="img" viewBox="0 0 512 512" style="height:1em;width:1em;vertical-align:-0.125em;margin-left:auto;margin-right:auto;font-size:inherit;fill:steelblue;overflow:visible;position:relative;"><path d="M459.37 151.716c.325 4.548.325 9.097.325 13.645 0 138.72-105.583 298.558-298.558 298.558-59.452 0-114.68-17.219-161.137-47.106 8.447.974 16.568 1.299 25.34 1.299 49.055 0 94.213-16.568 130.274-44.832-46.132-.975-84.792-31.188-98.112-72.772 6.498.974 12.995 1.624 19.818 1.624 9.421 0 18.843-1.3 27.614-3.573-48.081-9.747-84.143-51.98-84.143-102.985v-1.299c13.969 7.797 30.214 12.67 47.431 13.319-28.264-18.843-46.781-51.005-46.781-87.391 0-19.492 5.197-37.36 14.294-52.954 51.655 63.675 129.3 105.258 216.365 109.807-1.624-7.797-2.599-15.918-2.599-24.04 0-57.828 46.782-104.934 104.934-104.934 30.213 0 57.502 12.67 76.67 33.137 23.715-4.548 46.456-13.32 66.599-25.34-7.798 24.366-24.366 44.833-46.132 57.827 21.117-2.273 41.584-8.122 60.426-16.243-14.292 20.791-32.161 39.308-52.628 54.253z"/></svg> <span class="citation">@ClausEkstrom</span> ] --- class: animated, fadeIn layout: true ??? This slide defines a default animated fade-in slide transition --- background-image: url("pics/r2d2.jpeg") background-size: 100% --- # Causal guidance Need a DAG (and assumptions) for causal inference. <div class="grViz html-widget html-fill-item-overflow-hidden html-fill-item" id="htmlwidget-3c7ea42ded69edf83011" style="width:960px;height:280px;"></div> <script type="application/json" data-for="htmlwidget-3c7ea42ded69edf83011">{"x":{"diagram":"\n digraph {\n bgcolor = \"transparent\"\n rankdir = LR\n node [shape = none, fontcolor = \"white\", color = \"white\"]\n edge [color = \"white\"]\n Rain -> {Umbrella Outside}\n Temperature -> { Rain Umbrella }\n Outside -> {Umbrella}\n Rain[fontcolor=\"yellow\"]\n Umbrella[fontcolor=\"yellow\"]\n Temperature[fontcolor=\"gray\"]\n \n }","config":{"engine":"dot","options":null}},"evals":[],"jsHooks":[]}</script> Requires someone with .yellow[knowledge and understanding] about the field to propose relevant and realistic DAGs. --- class: middle, center background-image: url(pics/chat-start3.png) background-size: contain --- class: middle, center background-image: url(pics/chat-res.png) background-size: contain --- # Estimating causal effects with AI `$$\mathbb{E}(Y | A, X) = f(A, X)$$` for some functional form `\(f\)`. --- # Estimating causal effects with AI `$$\mathbb{E}(Y | A, X) = \text{AI}(A, X)$$` for some .yellow[AI]-model. Average Treatment Effect (ATE) `$$\mathbb{E}_X [ \underbrace{\mathbb{E}(Y | A=1, X)}_\text{Treated} - \underbrace{\mathbb{E}(Y | A=0, X)}_\text{Untreated}]$$` with estimator `\(\frac{1}{N}\sum_{i=1}^N [ \widehat{\mathbb{E}}(Y | A=1, X_i) - \widehat{\mathbb{E}}(Y | A=0, X_i)]\)` --- # AI `\(g\)`-formula algorithm 1. Estimate `\(\mathbb{E}(Y | A, X)\)` using an AI tool. 2. Set `\(A=1\)` for all observations and predict outcomes for all 3. Set `\(A=0\)` for all observations and predict outcomes for all `$$\frac{1}{N}\sum_{i=1}^N [ \widetilde{\mathbb{E}}(Y | A=1, X_i) - \widetilde{\mathbb{E}}(Y | A=0, X_i)]$$` To interpret *causally* (average causal treatment effect) we still need the standard causal assumptions *and* proper models. --- # Causal discovery / structure learning .pull-left[ *Let a DAG be given ...* Can we learn (parts of) the DAG from data? ] .pull-right[ <div class="grViz html-widget html-fill-item-overflow-hidden html-fill-item" id="htmlwidget-ef3f7cb87b7c80fd8218" style="width:460px;height:300px;"></div> <script type="application/json" data-for="htmlwidget-ef3f7cb87b7c80fd8218">{"x":{"diagram":"\n digraph {\n bgcolor = \"transparent\"\n layout = twopi\n node [shape = circle, fontcolor = \"white\", color = \"white\"]\n edge [color = \"white\"]\n A\n Y\n M\n X\n }","config":{"engine":"dot","options":null}},"evals":[],"jsHooks":[]}</script> ] -- The PC algorithm identifies *conditional independencies* among the variables. Nice properties - in the oracle setting. .caption-right-vertical[Spirtes & Glymour (1991). *An algorithm for fast recovery of sparse causal graphs*. Social Science Computer Review.] --- # Causal discovery / structure learning There is an edge `\(A − Y\)` if and only if `\(A\)` and `\(Y\)` are *dependent* conditional on *every possible subset* of the other variables. `\(A \perp Y\)`? <br> `\(A \perp Y | X\)`? <br> `\(A \perp Y | M\)`? <br> `\(A \perp Y | X, M\)`? All these tests need a model. Use .yellow[AI] to identify relevant tests for (conditional) independence testing? --- # Remember The premise for .yellow[artificial intelligence] is ... -- .pull-right[ ... .yellow[intelligence]. ] --- # References .small[ **Causal structure learning** Spirtes & Glymour (1991). *An algorithm for fast recovery of sparse causal graphs*. Social Science Computer Review. Petersen, Osler & Ekstrøm (2021). *Data-Driven Model Building for Life Course Epidemiology*. Am J Epidemiology. **Comparing DAGS from experts and machines** Petersen, Ekstrøm, Spirtes & Merete Osler (2023). *Constructing causal life course models: Comparative study of data-driven and theory-driven approaches*. Am J Epidemiology ]